Page 26 of 180
Posted: 24 Aug 2006, 12:45
by emile
conan_the_cribber wrote:
At the end of the last football season, Liverpool were the hottest team in England. If there had been predictive ratings for football, then at the end of the season Liverpool would've been ahead of Chelsea, because that would've helped in the task of obtaining the best percentage in predicting the results of the next matchups. However the 'new' career performance ratings reflects the performance over a longer period of time e.g. Chelsea ahead of Liverpool in the premiership table. As we've found out, most people want to see the league table, not the "whos in form list".
Good thing you don't follow American sports :)
Posted: 24 Aug 2006, 12:47
by conan_the_cribber
JCS83MD wrote:Conan,
I follow your argument on DQ. That is why, as you'll see in my latest attempts, I made DQ a 0.3, and split fight factors into 3 categories. Again, this is just testing I am not sure will ever be honored but when the stats improve, one would have to assume the change made better results than the programming that was once there.
I was planning on separating blowout decisions into its own category even higher than stoppages, however I assume this will negatively affect prediction rates and lose some clarity in the grand scheme of things. Also, it can be assumed that a fighter who went 12 rounds was not beat as badly as one who was knocked out, regardless of round or situation before this knockout. This argument has 2 valid sides to it.
What's unfortunate is that there is no real metric to gauge progress that is efficient and easy to interpret. Perhaps you, as a statistician, should provide one. :)
Hi JCS,
I'm afraid there is no silver bullet here.
If you read the Liverpool Chelsea analogy above, you will see where the problem lies with using the 'predictability' metric.
The predictability metric can only be used as a base statistic, to make sure there's nothing completely knucklehead about the system e.g. all Ukranian fighters at number 1.
The only solution I see, is to try to objective rank and rate fighters manually and see if that matches up to the ratings provided. For example I did this at one point in the discussion about light heavies. It made no sense to rank Briggs after Drews, because Drews had done nothing significant since losing. Erdei is perceived as the weakest of the four champs, can that be determined by his lack of KO's. Does that make the number of KO's a factor in peoples minds when ranking fighters.
So basically it's hard work. A group of people have to come with a plausible reason for ranking a group of fighters in a division (based only on results). There may be discrepensies e.g. Brock before Toney, Brock after Toney, but you have to be able to defend your position based solely on the info in boxrec, as if you had never seen the fight.
You then take the average of the results and match that up with what boxrec generates. If it fits, you know you're doing the right thing.
conan
Posted: 24 Aug 2006, 13:01
by JCS
conan_the_cribber wrote:JCS83MD wrote:Conan,
I follow your argument on DQ. That is why, as you'll see in my latest attempts, I made DQ a 0.3, and split fight factors into 3 categories. Again, this is just testing I am not sure will ever be honored but when the stats improve, one would have to assume the change made better results than the programming that was once there.
I was planning on separating blowout decisions into its own category even higher than stoppages, however I assume this will negatively affect prediction rates and lose some clarity in the grand scheme of things. Also, it can be assumed that a fighter who went 12 rounds was not beat as badly as one who was knocked out, regardless of round or situation before this knockout. This argument has 2 valid sides to it.
What's unfortunate is that there is no real metric to gauge progress that is efficient and easy to interpret. Perhaps you, as a statistician, should provide one. :)
Hi JCS,
I'm afraid there is no silver bullet here.
If you read the Liverpool Chelsea analogy above, you will see where the problem lies with using the 'predictability' metric.
The predictability metric can only be used as a base statistic, to make sure there's nothing completely knucklehead about the system e.g. all Ukranian fighters at number 1.
The only solution I see, is to try to objective rank and rate fighters manually and see if that matches up to the ratings provided. For example I did this at one point in the discussion about light heavies. It made no sense to rank Briggs after Drews, because Drews had done nothing significant since losing. Erdei is perceived as the weakest of the four champs, can that be determined by his lack of KO's. Does that make the number of KO's a factor in peoples minds when ranking fighters.
So basically it's hard work. A group of people have to come with a plausible reason for ranking a group of fighters in a division (based only on results). There may be discrepensies e.g. Brock before Toney, Brock after Toney, but you have to be able to defend your position based solely on the info in boxrec, as if you had never seen the fight.
You then take the average of the results and match that up with what boxrec generates. If it fits, you know you're doing the right thing.
conan
There really is no way to make the whole computer thing work. I strongly believe the IBO system has manual tweaking going on behind closed doors. Perfect evidence of this could be seen after Kirk Johnson lost to Mora and was still ranked above him, even though in all other similar situations I've seen, the winner was ranked above the loser. That's not to say this system is the best, or better than BoxRec's.. even though posters like Evander have wet dreams about it.
As you see, my latest update elevated Drews over Briggs, but it also dropped Jones below Erdei. This is the constant good/bad fight that will keep going on because of inconsistencies in judges scoring, varying hometown factors, etc., etc., the variables are endless and the editors cannot possibly enter all of them as some would merely be opinion. A human-managed system IS required for near-perfection. That's really all there is to it.
Predictability was the only truely objective thing to rely on WHEN the premise was entirely focused around prediction of course.
It is NOW my belief that in a computerized system, you need to have some niche that you can prove objectively (pure prediction rate) or you need a human there to enter subjective information and consistently enter all valid variables to have a chance at providing something worthwhile.
Posted: 24 Aug 2006, 14:43
by Lennox
JCS83MD wrote:conan_the_cribber wrote:JCS83MD wrote:Conan,
I follow your argument on DQ. That is why, as you'll see in my latest attempts, I made DQ a 0.3, and split fight factors into 3 categories. Again, this is just testing I am not sure will ever be honored but when the stats improve, one would have to assume the change made better results than the programming that was once there.
I was planning on separating blowout decisions into its own category even higher than stoppages, however I assume this will negatively affect prediction rates and lose some clarity in the grand scheme of things. Also, it can be assumed that a fighter who went 12 rounds was not beat as badly as one who was knocked out, regardless of round or situation before this knockout. This argument has 2 valid sides to it.
What's unfortunate is that there is no real metric to gauge progress that is efficient and easy to interpret. Perhaps you, as a statistician, should provide one. :)
Hi JCS,
I'm afraid there is no silver bullet here.
If you read the Liverpool Chelsea analogy above, you will see where the problem lies with using the 'predictability' metric.
The predictability metric can only be used as a base statistic, to make sure there's nothing completely knucklehead about the system e.g. all Ukranian fighters at number 1.
The only solution I see, is to try to objective rank and rate fighters manually and see if that matches up to the ratings provided. For example I did this at one point in the discussion about light heavies. It made no sense to rank Briggs after Drews, because Drews had done nothing significant since losing. Erdei is perceived as the weakest of the four champs, can that be determined by his lack of KO's. Does that make the number of KO's a factor in peoples minds when ranking fighters.
So basically it's hard work. A group of people have to come with a plausible reason for ranking a group of fighters in a division (based only on results). There may be discrepensies e.g. Brock before Toney, Brock after Toney, but you have to be able to defend your position based solely on the info in boxrec, as if you had never seen the fight.
You then take the average of the results and match that up with what boxrec generates. If it fits, you know you're doing the right thing.
conan
There really is no way to make the whole computer thing work. I strongly believe the IBO system has manual tweaking going on behind closed doors. Perfect evidence of this could be seen after Kirk Johnson lost to Mora and was still ranked above him, even though in all other similar situations I've seen, the winner was ranked above the loser. That's not to say this system is the best, or better than BoxRec's.. even though posters like Evander have wet dreams about it.
As you see, my latest update elevated Drews over Briggs, but it also dropped Jones below Erdei. This is the constant good/bad fight that will keep going on because of inconsistencies in judges scoring, varying hometown factors, etc., etc., the variables are endless and the editors cannot possibly enter all of them as some would merely be opinion. A human-managed system IS required for near-perfection. That's really all there is to it.
Predictability was the only truely objective thing to rely on WHEN the premise was entirely focused around prediction of course.
It is NOW my belief that in a computerized system, you need to have some niche that you can prove objectively (pure prediction rate) or you need a human there to enter subjective information and consistently enter all valid variables to have a chance at providing something worthwhile.
I dont think the IBO make manual changes JCS, if they did why are some of their champions rated at number 16 or whatever. They have always displayed things as it is. I have spoken with Ed Levine many times and I have told him about this forum, I know he looks in occasionally. I asked him about the Kirk Johnson situation and he said it was treated as an injury loss and the calculation was different. I have always found Ed Levine very honest, I dont see what they would gain by manually fixing them, Ive never seen anyone moved to number 50 just so they can make a contest. This thing about 'they wont tell me the formula so its no good' is irrelevant, you ring Coca Cola and ask for their recipe.
On another note as part of the ratings, would it be possible for Boxrec to include in the pull down not only, current, all time, but also WBC, WBA, IBF, IBO and WBO. it would enhance the site I think. Certainly useful when your trying to make a match.
The silver bullet
Posted: 24 Aug 2006, 17:55
by conan_the_cribber
The silver bullet....
10 ten bowling. A few beers, a load of trash talk, checking out chicks in the next lanes, basically a place where I do some of my best thinking. If I wanted to statistically judge if my algorithm was working, then this is how I'd do it.
The premise of the ratings is, that the more points a boxer has, the more they have proved in their career. As we all know, when a more experience boxer meets a lesser experienced boxer, then this doesn't necessarily translate to a win, or even the expectation of a win. There could be someone undefeated hot prospect coming up, meeting someone who was at the top but now on their way out.
However the following slightly more refined statement is more likely to be true. There are two fighters A and B. Fighter A has fought and beat a bunch of fighters above 300 in some recent time frame and not lost to anyone under 300. B has fought and lost to fighters lower than 300 and not beaten anyone above 300. If there was a match up between A and B, then I would expect A to win. Generalized, if A is currently beating fighters that B has recently lost against, then I expect A to beat B.
And more importantly, I would say, the measure of my confidence should be related to the current difference in the rankings. It doesn't have to be linear, but it should be increasing as the difference in rating increases. Why is this more important? It is more important because it helps you decide whether the way you are rating is correct or not. If I find no real difference in victory percentage when the difference is bigger, then that means I'm advancing people too soon. If I'm finding that there are less upsets when people are close, then I'm I might not be awarding the up and coming fighter enough points.
I would then collect a bunch of statistics across all fights where I have reason to believe that one fighter should win. I would probably group these by every 25 points or so. My expection from such a group of statistics, should be that it shows, that for every increase in the 25 point range, that the percentage of an upset gets less and less.
My real life current expectation is that here is probably not much difference in a 700 point difference to a 500 point difference i.e. the top fighters are being promoted too high. That's just a gut feeling, but it might be proved wrong.
Implementation details tomorrow.
conan
Posted: 24 Aug 2006, 19:19
by JCS
Woo farking hoo.. Less complaining, more action. That's what I like to see.
Posted: 25 Aug 2006, 04:44
by conan_the_cribber
JCS83MD wrote:Woo farking hoo.. Less complaining, more action. That's what I like to see.
JCS,
The most important action was to complain. Actions to improve the predictive ratings were futile, because the underlying concept was flawed.
conan
Posted: 25 Aug 2006, 08:46
by JCS
conan_the_cribber wrote:JCS83MD wrote:Woo farking hoo.. Less complaining, more action. That's what I like to see.
JCS,
The most important action was to complain. Actions to improve the predictive ratings were futile, because the underlying concept was flawed.
conan
Everyone's got an opinion.
Posted: 25 Aug 2006, 09:08
by emile
And again?
Posted: 25 Aug 2006, 09:11
by JCS
emile wrote:And again?
Hey you all asked for it. Looks like it was changed back to 2 days ago.
Re: I cant follow this
Posted: 25 Aug 2006, 10:18
by JCS
Cobwebcat wrote:Why is Acelino Freitas bouncing up and down the ratings for no apparent reason? 3rd 1st now 3rd again?
Conan: that seems a good way of checking things out to me.
I really liked the all time rankings I hope they are sorted out when computerrank gets back, and not to how JCS has them now (I'm not having a go at you here JCS I know you dont like them either)
Is it possible to keep current along the lines we have it now but keep predictive for the all-time? I know that sounds like a contradiction but when you are comparing fighters from different eras that's exactly what you are doing...predicting
What does everyone think?
I think the predictive system probably had the better all-time rankings, but I should try it considering 7 years, as opposed to 5. I think some guys got too high by just having a few good years.. Don't you think?
Posted: 25 Aug 2006, 10:28
by JCS
Cobwebcat wrote:Yeah worth a try. I've sent Computerrank a private post to ask to keep predictive for the all time. Hopefully we can keep the benefit of a lot of work for that rating without losing the adjustments made to Conan's current system.
I think he's gone on holiday though.
I'm not sure the current system can be considered Conan's hehe...
As for keeping both systems in tact for different purposes.. I doubt that's going to happen.. just my opinion.
I'll try running the best prediction ratings we have based on 7 best years. Hopefully Hamed won't be the Top Feather anymore.
Re: The silver bullet
Posted: 25 Aug 2006, 10:34
by JCS
conan_the_cribber wrote:The silver bullet....
10 ten bowling. A few beers, a load of trash talk, checking out chicks in the next lanes, basically a place where I do some of my best thinking. If I wanted to statistically judge if my algorithm was working, then this is how I'd do it.
The premise of the ratings is, that the more points a boxer has, the more they have proved in their career. As we all know, when a more experience boxer meets a lesser experienced boxer, then this doesn't necessarily translate to a win, or even the expectation of a win. There could be someone undefeated hot prospect coming up, meeting someone who was at the top but now on their way out.
However the following slightly more refined statement is more likely to be true. There are two fighters A and B. Fighter A has fought and beat a bunch of fighters above 300 in some recent time frame and not lost to anyone under 300. B has fought and lost to fighters lower than 300 and not beaten anyone above 300. If there was a match up between A and B, then I would expect A to win. Generalized, if A is currently beating fighters that B has recently lost against, then I expect A to beat B.
And more importantly, I would say, the measure of my confidence should be related to the current difference in the rankings. It doesn't have to be linear, but it should be increasing as the difference in rating increases. Why is this more important? It is more important because it helps you decide whether the way you are rating is correct or not. If I find no real difference in victory percentage when the difference is bigger, then that means I'm advancing people too soon. If I'm finding that there are less upsets when people are close, then I'm I might not be awarding the up and coming fighter enough points.
I would then collect a bunch of statistics across all fights where I have reason to believe that one fighter should win. I would probably group these by every 25 points or so. My expection from such a group of statistics, should be that it shows, that for every increase in the 25 point range, that the percentage of an upset gets less and less.
My real life current expectation is that here is probably not much difference in a 700 point difference to a 500 point difference i.e. the top fighters are being promoted too high. That's just a gut feeling, but it might be proved wrong.
Implementation details tomorrow.
conan
I don't think Martin believes in using a derived measurement in a verification or statistical metric.
Posted: 25 Aug 2006, 10:41
by JCS
JCS83MD wrote:CobWeb
Here is the Top 50 from the Best Predictive Ratings which never got released (5 Best Years)
Code: Select all
| Oscar | De La Hoya | 2064.81232134621 |
| Sugar Ray | Robinson | 2042.02613381611 |
| Ike | Williams | 2041.76700302035 |
| Muhammad | Ali | 2012.23727569841 |
| Henry | Armstrong | 2007.41491262254 |
| Carlos | Monzon | 2005.87586314998 |
| Naseem | Hamed | 2004.39359922628 |
| Marvin | Hagler | 2000.61655071512 |
| Floyd | Mayweather Jr | 1987.24941185935 |
| Sugar Ray | Leonard | 1987.15469869271 |
| Joe | Louis | 1986.44425411519 |
| Davey | Moore | 1984.18276636864 |
| Carlos | Ortiz | 1984.11591087347 |
| Mike | Tyson | 1977.16962033522 |
| Kid | Gavilan | 1962.94740893949 |
| Ezzard | Charles | 1940.88979491201 |
| Rocky | Marciano | 1939.52535915137 |
| Vicente | Saldivar | 1935.52581959475 |
| Jose Angel | Napoles | 1932.71622187618 |
| Pernell | Whitaker | 1931.22293703604 |
| Alexis | Arguello | 1929.01963460997 |
| Luis Manuel | Rodriguez | 1927.90875735805 |
| Roy | Jones Jr | 1927.5474157664 |
| Barney | Ross | 1927.53769331541 |
| Thomas | Hearns | 1925.10542466472 |
| Michael | Nunn | 1925.04864259831 |
| Kenny | Lane | 1921.4138192183 |
| Bernard | Hopkins | 1921.15491348588 |
| Larry | Holmes | 1916.87557701921 |
| Emile | Griffith | 1914.25771496534 |
| Archie | Moore | 1911.13931631247 |
| Aaron | Pryor | 1908.31390499913 |
| Felix | Trinidad | 1906.00325153051 |
| Kostya | Tszyu | 1903.54800569117 |
| Tony | Canzoneri | 1903.04666782354 |
| Carl 'Bobo' | Olson | 1900.85423690023 |
| Carmen | Basilio | 1898.63198606611 |
| Sugar | Ramos | 1898.27951960378 |
| Eder | Jofre | 1896.70154338189 |
| Beau | Jack | 1896.07488569562 |
| Sonny | Liston | 1894.94712086117 |
| Terry | Norris | 1893.60815365246 |
| Julio Cesar | Chavez | 1892.6948081644 |
| Joe | Frazier | 1882.79969733435 |
| Hector | Camacho | 1881.58767523425 |
| Marco Antonio | Barrera | 1879.78790301903 |
| Dick | Tiger | 1878.24503986678 |
| Eusebio | Pedroza | 1877.18842889006 |
| Gene | Fullmer | 1872.48208284744 |
| Willie | Pep | 1868.99568290903 |
Same system, 7 best years considered.. Surprisingly not much different
Top 50
Code: Select all
| Oscar | De La Hoya | 2013.26941012421 |
| Sugar Ray | Robinson | 1998.2789621562 |
| Ike | Williams | 1996.62352923294 |
| Muhammad | Ali | 1990.68411546553 |
| Marvin | Hagler | 1987.48246432334 |
| Floyd | Mayweather Jr | 1975.40076793952 |
| Carlos | Monzon | 1966.35920387298 |
| Carlos | Ortiz | 1963.95119127143 |
| Joe | Louis | 1961.91193665612 |
| Naseem | Hamed | 1956.77251028716 |
| Henry | Armstrong | 1945.92276107009 |
| Sugar Ray | Leonard | 1927.65362180083 |
| Davey | Moore | 1921.32828415341 |
| Roy | Jones Jr | 1918.73309043321 |
| Ezzard | Charles | 1918.7306885379 |
| Mike | Tyson | 1917.56856827231 |
| Luis Manuel | Rodriguez | 1917.49755836676 |
| Jose Angel | Napoles | 1911.86302649716 |
| Alexis | Arguello | 1909.4670934582 |
| Pernell | Whitaker | 1909.25797111012 |
| Kid | Gavilan | 1905.71525800068 |
| Thomas | Hearns | 1903.00367423432 |
| Larry | Holmes | 1895.17664052924 |
| Bernard | Hopkins | 1893.78783923653 |
| Archie | Moore | 1889.8348612258 |
| Barney | Ross | 1888.86454957819 |
| Felix | Trinidad | 1888.45069503482 |
| Emile | Griffith | 1888.36314390671 |
| Tony | Canzoneri | 1886.20925409356 |
| Kostya | Tszyu | 1884.57795703138 |
| Kenny | Lane | 1881.43362457623 |
| Rocky | Marciano | 1880.59731716777 |
| Michael | Nunn | 1879.96369712643 |
| Vicente | Saldivar | 1875.64928626061 |
| Julio Cesar | Chavez | 1875.30641341415 |
| Marco Antonio | Barrera | 1862.74715799856 |
| Terry | Norris | 1857.80648975969 |
| Hector | Camacho | 1849.88589714799 |
| Dick | Tiger | 1846.61863240212 |
| Evander | Holyfield | 1845.8112548062 |
| Sonny | Liston | 1844.73277066509 |
| Aaron | Pryor | 1844.41202996307 |
| Eusebio | Pedroza | 1844.00432854393 |
| Jimmy | McLarnin | 1841.88211495055 |
| Joe | Frazier | 1841.38100687785 |
| Beau | Jack | 1840.4137030849 |
| Eder | Jofre | 1839.8610815216 |
| Carl 'Bobo' | Olson | 1838.80489480782 |
| Sugar | Ramos | 1838.38247936403 |
| Carmen | Basilio | 1835.17299056967 |
Maybe I'll try 10...
Posted: 25 Aug 2006, 11:03
by JCS
10 Best Years AllTime - Best Prediction Ratings
I think this is a little more like it -- Of course, the newer fighters are even higher now that the problem with missing older bouts is amplified.
Code: Select all
| Muhammad | Ali | 1968.1251625707 |
| Oscar | De La Hoya | 1959.28572614872 |
| Sugar Ray | Robinson | 1935.81153739486 |
| Marvin | Hagler | 1933.10842081975 |
| Carlos | Ortiz | 1923.96464216368 |
| Joe | Louis | 1923.2089639969 |
| Roy | Jones Jr | 1904.18424793757 |
| Henry | Armstrong | 1887.71740973516 |
| Carlos | Monzon | 1885.65305785257 |
| Luis Manuel | Rodriguez | 1882.36790188951 |
| Ike | Williams | 1880.18040886633 |
| Pernell | Whitaker | 1869.82512280545 |
| Jose Angel | Napoles | 1868.65580312775 |
| Alexis | Arguello | 1865.54040589887 |
| Archie | Moore | 1864.45260084183 |
| Thomas | Hearns | 1861.9683447288 |
| Bernard | Hopkins | 1857.20430511866 |
| Kostya | Tszyu | 1855.68269126823 |
| Julio Cesar | Chavez | 1855.08073152515 |
| Sugar Ray | Leonard | 1854.94697960783 |
| Felix | Trinidad | 1854.828216085 |
| Larry | Holmes | 1853.79274492675 |
| Emile | Griffith | 1853.51082563689 |
| Ezzard | Charles | 1842.16350348961 |
| Mike | Tyson | 1840.58007683153 |
| Marco Antonio | Barrera | 1832.81146765555 |
| Tony | Canzoneri | 1828.45639909366 |
| Kenny | Lane | 1823.24544210956 |
| Kid | Gavilan | 1820.23633998656 |
| Evander | Holyfield | 1816.43737376433 |
| Erik | Morales | 1816.02494297796 |
| Lennox | Lewis | 1805.73888401199 |
| Dick | Tiger | 1801.85349125785 |
| Jimmy | McLarnin | 1801.57870364853 |
| Davey | Moore | 1800.09782615457 |
| Michael | Nunn | 1799.30429388925 |
| Hector | Camacho | 1799.14755406457 |
| Eddie | Perkins | 1796.94136732255 |
| Ken | Buchanan | 1794.00792739424 |
| Terry | Norris | 1793.26302655351 |
| Azumah | Nelson | 1792.56000743617 |
| Young | Corbett III | 1789.63308373352 |
| Sonny | Liston | 1789.02416137007 |
| Roberto | Duran | 1789.01105432088 |
| Joe | Frazier | 1788.57601882586 |
| Brian | Mitchell | 1783.76002765919 |
| Ralph | Dupas | 1782.52051519799 |
| Vicente | Saldivar | 1782.04687383585 |
| Willie | Pep | 1779.27200256122 |
| Sugar | Ramos | 1776.09589437539 |
Posted: 25 Aug 2006, 11:14
by JCS
Cobwebcat wrote:Agreed though PBF going from top ten to nowhere seems a step to far. Perhaps 8 yrs?
Perhaps.. One could argue he hasn't fought quite long enough yet though.. Him not being in the Top 10 yet is definitely valid, but you could argue him in the Top 50.. I'll try 9 -- BTW, Floyd was #94 using 10 years.
But even though we're improving results.. All the worry at the top because of constant complaining tends to dull any achievements which aren't consider "proper" to those traditionalists.
Posted: 25 Aug 2006, 11:37
by JCS
Cobwebcat wrote:I dont think you would get so much complaining with the all time. Particlarlyas they even "look better"
8 looks nearly identical to 7.. I'll try 9, but I think 10 is the winner.. Its a nice even number, and I think a fighter should have to put this many quality years in.
IT should probably even be modified to weight the top year more so than the others.. but thats just my opinion.
Posted: 25 Aug 2006, 11:56
by JCS
CobWebCat wrote:JCS83MD wrote:IT should probably even be modified to weight the top year more so than the others.. but thats just my opinion.
I definietly agree with that.
If we used 8 years for example I'd go:
Best 2 years 40% of total points
Next two best years 30% of total
Years 5 and 6 20% of total
Years 7 and 8 10% of total
That would be interestig to see.
9 looked very similar to 7 and 8.. You don't really start getting effects until 10. Perhaps I'll try some weighting.
Posted: 25 Aug 2006, 12:41
by JCS
Ok, sorry for the delay, the ranking programming was somewhat tricky. Using the best prediction system.. I'm gonna go with 10 years
Code: Select all
1st Best = 16%
2nd Best = 14%
3rd Best = 12%
4th Best = 11%
5th Best = 10%
6th Best = 10%
7th Best = 9%
8th Best = 8%
9th Best = 6%
10th Best = 4%
| Oscar | De La Hoya | 1996.58811884424 |
| Muhammad | Ali | 1982.44075247179 |
| Sugar Ray | Robinson | 1979.36814368442 |
| Marvin | Hagler | 1959.20742778952 |
| Joe | Louis | 1950.13437784894 |
| Ike | Williams | 1946.64955649738 |
| Carlos | Ortiz | 1946.40317455478 |
| Henry | Armstrong | 1931.33549193201 |
| Carlos | Monzon | 1928.49432348768 |
| Roy | Jones Jr | 1913.89546534314 |
| Sugar Ray | Leonard | 1905.05433872475 |
| Luis Manuel | Rodriguez | 1902.95142649334 |
| Alexis | Arguello | 1895.22162802568 |
| Pernell | Whitaker | 1895.16834530582 |
| Jose Angel | Napoles | 1894.51172801839 |
| Mike | Tyson | 1889.80836064068 |
| Thomas | Hearns | 1887.33098341423 |
| Ezzard | Charles | 1883.79303299929 |
| Archie | Moore | 1883.0878677229 |
| Bernard | Hopkins | 1880.5607348495 |
| Emile | Griffith | 1878.09882253968 |
| Larry | Holmes | 1876.48194097491 |
| Felix | Trinidad | 1876.44648212334 |
| Kostya | Tszyu | 1875.76534402291 |
| Kid | Gavilan | 1875.73862892915 |
| Floyd | Mayweather Jr | 1870.21321566594 |
| Julio Cesar | Chavez | 1869.83061591091 |
| Davey | Moore | 1867.8322259063 |
| Kenny | Lane | 1859.93045023251 |
| Tony | Canzoneri | 1859.15477536967 |
| Naseem | Hamed | 1858.77512098503 |
| Marco Antonio | Barrera | 1852.5019413308 |
| Michael | Nunn | 1845.84280060719 |
| Vicente | Saldivar | 1836.84599637585 |
| Evander | Holyfield | 1836.6561808245 |
| Terry | Norris | 1836.12832431791 |
| Barney | Ross | 1832.64311572797 |
| Dick | Tiger | 1832.40116210528 |
| Hector | Camacho | 1828.46382961066 |
| Erik | Morales | 1826.48661280108 |
| Jimmy | McLarnin | 1826.25023223955 |
| Sonny | Liston | 1826.23557532254 |
| Joe | Frazier | 1821.09653347912 |
| Lennox | Lewis | 1821.03834460682 |
| Willie | Pep | 1820.81604804496 |
| Carl 'Bobo' | Olson | 1818.27659620463 |
| Ken | Buchanan | 1817.78793579847 |
| Sugar | Ramos | 1817.55606724941 |
| Eddie | Perkins | 1814.85209628291 |
| Azumah | Nelson | 1812.74672470151 |
widegap to the Ratings - No?
Posted: 31 Aug 2006, 15:27
by jimglen
wide gap to the Ratings - No?
please explain how there can be 100s and even 1000s of points between the Top men...is this not very much a misrepresentation to the lower Top men, in some cases beginning as 'early' as No.3 (?).
Posted: 03 Sep 2006, 11:18
by JCS
After careful deliberation, this is my main argument about the ratings.
There will never be a perfectly accurate objective traditional-based system where everyone agrees with the results. This problem is amplified by the fact that a computer, and a computer only, has the only access to the ratings (no manual tweaks). Everyone else has a traditional system, so this will result in further knocks against the computerized system, because it simply cannot interpret all the subtle factors that can go into ranking a fighter. As BoxBuzz said, we can all paint a picture on how the current rankings look, or should look, so why build a computerized system to do the same exact thing?
BoxRec is the best site on the net for boxing records, hands down, so why shouldn't it have a rating system that is the best at what it does? The predictive system on BoxRec was the best around. It relied on no subjectivity and there were no equivalents, so it was impossible to really put it down, other than to wait out the results. Woods vs. Johnson and Peter vs. Toney, were miraculously predicted correctly (even though I disagreed with both decisions).
Most of the grief that the predictive system received were from two sources:
1. Strict traditionalists who could not wrap their heads around what a predictive system was made to do, either because they are so mentally bound to a traditional system, or because they really could not understand it.
2. People stumbling upon the site without clue as to what our ratings were trying to accomplish.
#2 would've been easily solved by a more visible disclaimer. #1, well they could've consulted other sources like "The Ring." Why let the lower-functioning and/or hard-headed mentalities of that group spoil it for the ones who really understand?
Now, I truely believe you are better off consulting FightNews' ratings, or BoxingTalk's ratings because they can produce the subtleties in their traditional systems that only a human could interject with. The system here now is a big juggling game.
Posted: 03 Sep 2006, 15:11
by conan_the_cribber
JCS83MD wrote:After careful deliberation, this is my main argument about the ratings.
There will never be a perfectly accurate objective traditional-based system where everyone agrees with the results. This problem is amplified by the fact that a computer, and a computer only, has the only access to the ratings (no manual tweaks). Everyone else has a traditional system, so this will result in further knocks against the computerized system, because it simply cannot interpret all the subtle factors that can go into ranking a fighter. As BoxBuzz said, we can all paint a picture on how the current rankings look, or should look, so why build a computerized system to do the same exact thing?
BoxRec is the best site on the net for boxing records, hands down, so why shouldn't it have a rating system that is the best at what it does? The predictive system on BoxRec was the best around. It relied on no subjectivity and there were no equivalents, so it was impossible to really put it down, other than to wait out the results. Woods vs. Johnson and Peter vs. Toney, were miraculously predicted correctly (even though I disagreed with both decisions).
Most of the grief that the predictive system received were from two sources:
1. Strict traditionalists who could not wrap their heads around what a predictive system was made to do, either because they are so mentally bound to a traditional system, or because they really could not understand it.
2. People stumbling upon the site without clue as to what our ratings were trying to accomplish.
#2 would've been easily solved by a more visible disclaimer. #1, well they could've consulted other sources like "The Ring." Why let the lower-functioning and/or hard-headed mentalities of that group spoil it for the ones who really understand?
Now, I truely believe you are better off consulting FightNews' ratings, or BoxingTalk's ratings because they can produce the subtleties in their traditional systems that only a human could interject with. The system here now is a big juggling game.
The winner of Shannon Briggs vs Brian Minto should be the number one fighter! Undoubtably the best, super duper that it can possibly be achieved.
JCS, re-read the stuff about Liverpool and Chelsea. That is the essence of the problem. The predictive rankings were a form guide. So in form boxers got rated higher. With the above example of totally crap results or outliers as we called them.
As far as I'm concerned, publish the damm predictive as well, the univers e is big enough for both of them. Just dont make them the real rankings next to a boxers name, it's embarrassing.
conan
Posted: 03 Sep 2006, 15:42
by JCS
conan_the_cribber wrote:JCS83MD wrote:After careful deliberation, this is my main argument about the ratings.
There will never be a perfectly accurate objective traditional-based system where everyone agrees with the results. This problem is amplified by the fact that a computer, and a computer only, has the only access to the ratings (no manual tweaks). Everyone else has a traditional system, so this will result in further knocks against the computerized system, because it simply cannot interpret all the subtle factors that can go into ranking a fighter. As BoxBuzz said, we can all paint a picture on how the current rankings look, or should look, so why build a computerized system to do the same exact thing?
BoxRec is the best site on the net for boxing records, hands down, so why shouldn't it have a rating system that is the best at what it does? The predictive system on BoxRec was the best around. It relied on no subjectivity and there were no equivalents, so it was impossible to really put it down, other than to wait out the results. Woods vs. Johnson and Peter vs. Toney, were miraculously predicted correctly (even though I disagreed with both decisions).
Most of the grief that the predictive system received were from two sources:
1. Strict traditionalists who could not wrap their heads around what a predictive system was made to do, either because they are so mentally bound to a traditional system, or because they really could not understand it.
2. People stumbling upon the site without clue as to what our ratings were trying to accomplish.
#2 would've been easily solved by a more visible disclaimer. #1, well they could've consulted other sources like "The Ring." Why let the lower-functioning and/or hard-headed mentalities of that group spoil it for the ones who really understand?
Now, I truely believe you are better off consulting FightNews' ratings, or BoxingTalk's ratings because they can produce the subtleties in their traditional systems that only a human could interject with. The system here now is a big juggling game.
The winner of Shannon Briggs vs Brian Minto should be the number one fighter! Undoubtably the best, super duper that it can possibly be achieved.
JCS, re-read the stuff about Liverpool and Chelsea. That is the essence of the problem. The predictive rankings were a form guide. So in form boxers got rated higher. With the above example of totally crap results or outliers as we called them.
As far as I'm concerned, publish the damm predictive as well, the univers e is big enough for both of them. Just dont make them the real rankings next to a boxers name, it's embarrassing.
conan
Maybe a more refined metric to use as analysis would solve problems such as Briggs and Minto... but that's the beauty of that system, its flaws can be accepted as long as people understand the basis of that system. Since there's no Minto Briggs fight on the horizon.. its really a moot point. They could draw each other.
Should Roy Jones really be ahead of Zsolt Erdei?
Does Luis Collazo really not deserve a Top 40 ranking?
Harrison above JMM?
Definitely unacceptable considering a TRADITIONAL ranking..
Posted: 04 Sep 2006, 03:54
by conan_the_cribber
JCS83MD wrote:conan_the_cribber wrote:JCS83MD wrote:After careful deliberation, this is my main argument about the ratings.
There will never be a perfectly accurate objective traditional-based system where everyone agrees with the results. This problem is amplified by the fact that a computer, and a computer only, has the only access to the ratings (no manual tweaks). Everyone else has a traditional system, so this will result in further knocks against the computerized system, because it simply cannot interpret all the subtle factors that can go into ranking a fighter. As BoxBuzz said, we can all paint a picture on how the current rankings look, or should look, so why build a computerized system to do the same exact thing?
BoxRec is the best site on the net for boxing records, hands down, so why shouldn't it have a rating system that is the best at what it does? The predictive system on BoxRec was the best around. It relied on no subjectivity and there were no equivalents, so it was impossible to really put it down, other than to wait out the results. Woods vs. Johnson and Peter vs. Toney, were miraculously predicted correctly (even though I disagreed with both decisions).
Most of the grief that the predictive system received were from two sources:
1. Strict traditionalists who could not wrap their heads around what a predictive system was made to do, either because they are so mentally bound to a traditional system, or because they really could not understand it.
2. People stumbling upon the site without clue as to what our ratings were trying to accomplish.
#2 would've been easily solved by a more visible disclaimer. #1, well they could've consulted other sources like "The Ring." Why let the lower-functioning and/or hard-headed mentalities of that group spoil it for the ones who really understand?
Now, I truely believe you are better off consulting FightNews' ratings, or BoxingTalk's ratings because they can produce the subtleties in their traditional systems that only a human could interject with. The system here now is a big juggling game.
The winner of Shannon Briggs vs Brian Minto should be the number one fighter! Undoubtably the best, super duper that it can possibly be achieved.
JCS, re-read the stuff about Liverpool and Chelsea. That is the essence of the problem. The predictive rankings were a form guide. So in form boxers got rated higher. With the above example of totally crap results or outliers as we called them.
As far as I'm concerned, publish the damm predictive as well, the univers e is big enough for both of them. Just dont make them the real rankings next to a boxers name, it's embarrassing.
conan
Maybe a more refined metric to use as analysis would solve problems such as Briggs and Minto... but that's the beauty of that system, its flaws can be accepted as long as people understand the basis of that system. Since there's no Minto Briggs fight on the horizon.. its really a moot point. They could draw each other.
Should Roy Jones really be ahead of Zsolt Erdei?
Does Luis Collazo really not deserve a Top 40 ranking?
Harrison above JMM?
Definitely unacceptable considering a TRADITIONAL ranking..
Well there you go, hoist on your own petard. You're arguing things that are totally borderline i.e. Jones ahead of Erdei (based entirely on results its totally justifiable) and I'm arguing things like Shannon friggen Briggs and minto as top 12 and Stipe drews at #2 at light heavy that are just red hot stinking turds in the world of rankings.
If that's all you got to compare, then you got no argument.
conan
Posted: 04 Sep 2006, 05:24
by Lennox
JCS83MD wrote:conan_the_cribber wrote:JCS83MD wrote:After careful deliberation, this is my main argument about the ratings.
There will never be a perfectly accurate objective traditional-based system where everyone agrees with the results. This problem is amplified by the fact that a computer, and a computer only, has the only access to the ratings (no manual tweaks). Everyone else has a traditional system, so this will result in further knocks against the computerized system, because it simply cannot interpret all the subtle factors that can go into ranking a fighter. As BoxBuzz said, we can all paint a picture on how the current rankings look, or should look, so why build a computerized system to do the same exact thing?
BoxRec is the best site on the net for boxing records, hands down, so why shouldn't it have a rating system that is the best at what it does? The predictive system on BoxRec was the best around. It relied on no subjectivity and there were no equivalents, so it was impossible to really put it down, other than to wait out the results. Woods vs. Johnson and Peter vs. Toney, were miraculously predicted correctly (even though I disagreed with both decisions).
Most of the grief that the predictive system received were from two sources:
1. Strict traditionalists who could not wrap their heads around what a predictive system was made to do, either because they are so mentally bound to a traditional system, or because they really could not understand it.
2. People stumbling upon the site without clue as to what our ratings were trying to accomplish.
#2 would've been easily solved by a more visible disclaimer. #1, well they could've consulted other sources like "The Ring." Why let the lower-functioning and/or hard-headed mentalities of that group spoil it for the ones who really understand?
Now, I truely believe you are better off consulting FightNews' ratings, or BoxingTalk's ratings because they can produce the subtleties in their traditional systems that only a human could interject with. The system here now is a big juggling game.
The winner of Shannon Briggs vs Brian Minto should be the number one fighter! Undoubtably the best, super duper that it can possibly be achieved.
JCS, re-read the stuff about Liverpool and Chelsea. That is the essence of the problem. The predictive rankings were a form guide. So in form boxers got rated higher. With the above example of totally crap results or outliers as we called them.
As far as I'm concerned, publish the damm predictive as well, the univers e is big enough for both of them. Just dont make them the real rankings next to a boxers name, it's embarrassing.
conan
Maybe a more refined metric to use as analysis would solve problems such as Briggs and Minto... but that's the beauty of that system, its flaws can be accepted as long as people understand the basis of that system. Since there's no Minto Briggs fight on the horizon.. its really a moot point. They could draw each other.
Should Roy Jones really be ahead of Zsolt Erdei?
Does Luis Collazo really not deserve a Top 40 ranking?
Harrison above JMM?
Definitely unacceptable considering a TRADITIONAL ranking..
I think the 'predictive thing' was ok as a form guide too, but you should not have called them rankings (or ratings). The best boxing rankings should largely have the champions at the top and the Calvin brocks, Chad Dawsons made them look totally unaceptable as 'rankings'. I dont think that if a set of ratings predicts better than another they are better and that is where you JCS seemed to be stuck that, that was the criteria to measure.
In boxing we have a young fighter who we match with fighters say to 12-0 before we move that fighter into regional /class B fghters/ 8 round class. Perhaps a fighter only gets tested properly at the 18-0 20-0 level, its pretty obvious Lou Duva, Olympians or pretty much any fighter signed to a TV promoter is gonna win his first 15 or so fights beause of careful matching, but that does not bean the 15-0 is better than some 8-11-2 fighter from Illinois, who fights anyone. The predictive ratings that were produced did not reflect the quality of some of the journeymen, who to me are the gatekeepers and the people that you have to beat to ascend the ratings. If a boxer is number 18, he should have beaten a fighter that is or has been no worse than say top 25. If a boxer is number 9 he should have beaten someone on the top 10, not mugged his way like Calvin Brock, perhaps being 30-0 but only really having faced one or two top 30 fighters. So as a set of ratings I feel the criteria you used reflected number of wins rather than quality. perhaps you could develop a hybrid where you had different method at say the top 10, top 20 level. i think the Boxrec ratings are the best they have been at the moment, but as you know I like the IBO and the Boxrec now is pretty much a copy.
When two boxers collide and a number 5 challenger fights the champ 'upsets' will occur, that is boxing. So for me predictive ratinngs is not the way to measure, If i have a fighter at number 85, I look for fighters between 45-75 who I am pretty sure my man will beat. My eyes and ears would be a better indicator of the winner than any set of ratings.