Page 39 of 180
Posted: 21 Jul 2007, 22:24
by emile
JCS wrote:Discovered a new inactive penalty formula.
Does anyone agree that the experimental method makes a bit more sense in determining how many rating points of a boxer to keep, after inactivity?
Now, only 50% of a boxer's rating points are kept if he were to come back after 18 months of inactivity. Had he come back 1 day sooner, ALL of his points would remain...
I think your curve is better for sure. But I think it would be better still to not have a big shelf at all. If you could find an appropriate starting point and
have a very shallow curve that added a point daily or weekly. A sudden plunge could be gamed, if the rating formula was used for some official purpose.
Posted: 22 Jul 2007, 00:35
by JCS
emile wrote:JCS wrote:Discovered a new inactive penalty formula.
Does anyone agree that the experimental method makes a bit more sense in determining how many rating points of a boxer to keep, after inactivity?
Now, only 50% of a boxer's rating points are kept if he were to come back after 18 months of inactivity. Had he come back 1 day sooner, ALL of his points would remain...
I think your curve is better for sure. But I think it would be better still to not have a big shelf at all. If you could find an appropriate starting point and
have a very shallow curve that added a point daily or weekly. A sudden plunge could be gamed, if the rating formula was used for some official purpose.
I agree... but a small shelf is better than a large one.
I am trying to convince Martin of this, but I cannot succeed.....
Posted: 24 Jul 2007, 10:48
by JCS
Perhaps when Vitali comes back, pounces McCline, and still isn't even in the Top 10, or 12 maybe.. some people will take notice.
Posted: 25 Jul 2007, 09:14
by computerrank
... McCline is #42, so this would not help very much to approve a top 10 rank.
Posted: 25 Jul 2007, 09:21
by JCS
computerrank wrote:... McCline is #42, so this would not help very much to approve a top 10 rank.
But Vitali was likely #1 or 2 when he retired.. so to see him so low when he comes back will be disheartening.
Like emile said, the large shelf in the current penalty is far too severe.
Posted: 25 Jul 2007, 10:49
by m1kee50
why not set it at .1% for each month off? so 10 month layoff (not that unusual for a world champ these days = 1% penalty
when they reach a years layoff they are found to be inactive by the boxrec ratings anyway, you could tack a fixed penalty per year onto it, and keep or discard the .1% a month as the maths allows it
Posted: 25 Jul 2007, 10:53
by JCS
~ MIKEE ~ wrote:why not set it at .1% for each month off? so 10 month layoff (not that unusual for a world champ these days = 1% penalty
when they reach a years layoff they are found to be inactive by the boxrec ratings anyway, you could tack a fixed penalty per year onto it, and keep or discard the .1% a month as the maths allows it
That's how the old prediction system was.. I think it was 0.011% per day... so something like .35% per month.
I think for a traditional system, we probably wouldn't wanna start it until 12 months or so, but yeah.. I agree, something should be done.
Posted: 25 Jul 2007, 10:59
by m1kee50
JCS wrote:~ MIKEE ~ wrote:why not set it at .1% for each month off? so 10 month layoff (not that unusual for a world champ these days = 1% penalty
when they reach a years layoff they are found to be inactive by the boxrec ratings anyway, you could tack a fixed penalty per year onto it, and keep or discard the .1% a month as the maths allows it
That's how the old prediction system was.. I think it was 0.011% per day... so something like .35% per month.
I think for a traditional system, we probably wouldn't wanna start it until 12 months or so, but yeah.. I agree, something should be done.
IC - well im sure you can code it for the effect you desire - I would have thought that a yearly penalty to punish long term inactivity would be good but maybe inflicted daily/weekly/monthly so as to be realistic - otherwise they get the 6-12 months 'free' (the 0-6 months period being a 'grace' period after the fight I assume)
Posted: 25 Jul 2007, 11:02
by JCS
~ MIKEE ~ wrote:JCS wrote:~ MIKEE ~ wrote:why not set it at .1% for each month off? so 10 month layoff (not that unusual for a world champ these days = 1% penalty
when they reach a years layoff they are found to be inactive by the boxrec ratings anyway, you could tack a fixed penalty per year onto it, and keep or discard the .1% a month as the maths allows it
That's how the old prediction system was.. I think it was 0.011% per day... so something like .35% per month.
I think for a traditional system, we probably wouldn't wanna start it until 12 months or so, but yeah.. I agree, something should be done.
IC - well im sure you can code it for the effect you desire - I would have thought that a yearly penalty to punish long term inactivity would be good but maybe inflicted daily/weekly/monthly so as to be realistic - otherwise they get the 6-12 months 'free' (the 0-6 months period being a 'grace' period after the fight I assume)
Yes. I agree.
Its not fair to say that a boxer who comes back after 17 months and 29 days of inactivity is not penalized, while the boxer who comes back after 18 months and 1 day gets 50% of their points taken away.
Posted: 25 Jul 2007, 11:04
by m1kee50
JCS wrote:~ MIKEE ~ wrote:JCS wrote:
That's how the old prediction system was.. I think it was 0.011% per day... so something like .35% per month.
I think for a traditional system, we probably wouldn't wanna start it until 12 months or so, but yeah.. I agree, something should be done.
IC - well im sure you can code it for the effect you desire - I would have thought that a yearly penalty to punish long term inactivity would be good but maybe inflicted daily/weekly/monthly so as to be realistic - otherwise they get the 6-12 months 'free' (the 0-6 months period being a 'grace' period after the fight I assume)
Yes. I agree.
Its not fair to say that a boxer who comes back after 17 months and 29 days of inactivity is not penalized, while the boxer who comes back after 18 months and 1 day gets 50% of their points taken away.
is that the current cutoff date? 18 months from last fight? that seems a long time.... I would have thought 6 months grace would have been cool - if the penalty was exponential based then you could set it up so that 18 months (legal or injury trouble) would nt
totally kill your rankings, but around 24 months it started biting
Posted: 25 Jul 2007, 11:06
by JCS
~ MIKEE ~ wrote:JCS wrote:~ MIKEE ~ wrote:
IC - well im sure you can code it for the effect you desire - I would have thought that a yearly penalty to punish long term inactivity would be good but maybe inflicted daily/weekly/monthly so as to be realistic - otherwise they get the 6-12 months 'free' (the 0-6 months period being a 'grace' period after the fight I assume)
Yes. I agree.
Its not fair to say that a boxer who comes back after 17 months and 29 days of inactivity is not penalized, while the boxer who comes back after 18 months and 1 day gets 50% of their points taken away.
is that the current cutoff date? 18 months from last fight? that seems a long time.... I would have thought 6 months grace would have been cool - if the penalty was exponential based then you could set it up so that 18 months (legal or injury trouble) would nt
totally kill your rankings, but around 24 months it started biting
Well, if you use a LARGE cut-off like we do now.. you have to be more than fair, otherwise you'll have crippled ratings everywhere. Even the experimental formula I tried with the pretty graph was a bit too steep to reduce from 18 months.
If we do a more gradual approach, I agree, it should be lowered.
Posted: 25 Jul 2007, 11:15
by m1kee50
JCS do the all time rankings mean that if Ricky Hatton retired today he would be the 7th greatest LWW of all-time? or is there a further calculation in that case?
Posted: 25 Jul 2007, 12:58
by m1kee50
~ MIKEE ~ wrote:JCS do the all time rankings mean that if Ricky Hatton retired today he would be the 7th greatest LWW of all-time? or is there a further calculation in that case?
also - what would be the obstacle to providing an option that the viewer could look at the ratings for a particular year? obviously at a certain date might be problematic, but averaging out the years rankings so that if we wanted to see what the ratings system thoguth of the LHW's in 1980 for example?
Posted: 25 Jul 2007, 13:07
by JCS
~ MIKEE ~ wrote:~ MIKEE ~ wrote:JCS do the all time rankings mean that if Ricky Hatton retired today he would be the 7th greatest LWW of all-time? or is there a further calculation in that case?
also - what would be the obstacle to providing an option that the viewer could look at the ratings for a particular year? obviously at a certain date might be problematic, but averaging out the years rankings so that if we wanted to see what the ratings system thoguth of the LHW's in 1980 for example?
If Ricky retired today, his all-time rating would be as it is now.
There already are year-by-year rankings in the BoxRec Encyclopedia, although I'm fairly certain they are based on an older revision of the rating system. The obstacle would be that there is no spot in the database to store these historical rankings unless JohnShep added this. It would be quite a bit of work for him I imagine.
Posted: 25 Jul 2007, 13:14
by computerrank
Posted: 25 Jul 2007, 13:15
by m1kee50
Posted: 25 Jul 2007, 14:30
by m1kee50
Looking at say Jack Johnsons profile, he won the world title, what title does that correspond to in the title list, i would like to look at the lineage
Posted: 26 Jul 2007, 14:02
by JCS
~ MIKEE ~ wrote:Looking at say Jack Johnsons profile, he won the world title, what title does that correspond to in the title list, i would like to look at the lineage
I don't know
Posted: 26 Jul 2007, 14:03
by m1kee50
Posted: 26 Jul 2007, 14:04
by JCS
I'm gonna point the finger at a bug.
Posted: 26 Jul 2007, 14:07
by m1kee50
JCS wrote:
I'm gonna point the finger at a bug.
ok, I now need a map or a dictionary, not sure which.... you mean a bug on the system?
Posted: 30 Jul 2007, 11:51
by JCS
Cobwebcat wrote:Did something happen to the ratings. P4P has altered quite a lot overnight?
Something to do with inactivity?
Just typical fluctuations.. and the ratings weren't updated Sunday.
The P4P list isn't really a P4P list. Its more like a division dominance list.
EDIT: Apparently there will be the database field to a hold a true P4P rating... I have to remind ol' Johnny boy to add it in.
Posted: 07 Aug 2007, 10:26
by JCS
Thought this was interesting... The highest rating ever attained, and its progression throughout boxing history.
Division which bout took place (If set) is in parantheses.
Post-Bout Ratings
Code: Select all
NEW RECORD on 1849-02-07 - Tom Hyer (Heavyweight) @ 25
NEW RECORD on 1853-10-12 - John Morrisey (Heavyweight) @ 27
NEW RECORD on 1860-02-10 - Joe Goss (Middleweight) @ 27
NEW RECORD on 1860-03-20 - Joe Goss (Middleweight) @ 31
NEW RECORD on 1860-07-17 - Joe Goss (Lightweight) @ 35
NEW RECORD on 1862-07-28 - Posh Price (Middleweight) @ 45
NEW RECORD on 1862-11-25 - Joe Goss (Middleweight) @ 63
NEW RECORD on 1879-06-01 - Larry Foley (Middleweight) @ 77
NEW RECORD on 1882-12-01 - (Professor) Charles Hadley () @ 94
NEW RECORD on 1882-12-07 - (Professor) Charles Hadley () @ 98
NEW RECORD on 1885-01-05 - Barney Donovan (Lightweight) @ 173
NEW RECORD on 1885-04-25 - Barney Donovan (Lightweight) @ 176
NEW RECORD on 1886-03-19 - Jack Burke (Bantamweight) @ 257
NEW RECORD on 1887-09-24 - Peter Jackson () @ 271
NEW RECORD on 1892-04-21 - Joe Choynski () @ 292
NEW RECORD on 1892-05-30 - Peter Jackson (Heavyweight) @ 321
NEW RECORD on 1897-12-17 - Charles Kid McCoy (Middleweight) @ 379
NEW RECORD on 1898-12-23 - Tommy Ryan () @ 387
NEW RECORD on 1899-02-22 - Tommy Ryan () @ 390
NEW RECORD on 1899-04-19 - Tommy Ryan () @ 393
NEW RECORD on 1899-08-31 - Tommy Ryan (Middleweight) @ 412
NEW RECORD on 1899-09-18 - Tommy Ryan (Middleweight) @ 414
NEW RECORD on 1900-01-12 - Charles Kid McCoy () @ 452
NEW RECORD on 1900-02-01 - Terry McGovern (Featherweight) @ 456
NEW RECORD on 1900-03-09 - Terry McGovern (Featherweight) @ 457
NEW RECORD on 1900-03-15 - Terry McGovern () @ 468
NEW RECORD on 1900-03-23 - Frank Erne (Lightweight) @ 511
NEW RECORD on 1900-07-16 - Terry McGovern () @ 669
NEW RECORD on 1900-11-13 - Terry McGovern () @ 669
NEW RECORD on 1900-12-13 - Terry McGovern (Lightweight) @ 704
NEW RECORD on 1901-04-30 - Terry McGovern (Featherweight) @ 808
NEW RECORD on 1903-12-11 - Jack Johnson () @ 871
NEW RECORD on 1904-04-22 - Jack Johnson () @ 889
NEW RECORD on 1906-09-03 - Joe Gans (Lightweight) @ 960
NEW RECORD on 1907-01-01 - Joe Gans (Lightweight) @ 1023
NEW RECORD on 1907-09-09 - Joe Gans (Lightweight) @ 1174
NEW RECORD on 1907-09-27 - Joe Gans (Lightweight) @ 1196
NEW RECORD on 1909-10-16 - Jack Johnson (Heavyweight) @ 1203
NEW RECORD on 1912-04-26 - Packey McFarland (Lightweight) @ 1271
NEW RECORD on 1912-08-03 - Sam Langford () @ 1405
NEW RECORD on 1912-10-09 - Sam Langford () @ 1483
NEW RECORD on 1912-12-26 - Sam Langford (Heavyweight) @ 1516
NEW RECORD on 1915-11-23 - Ted 'Kid' Lewis (Welterweight) @ 1644
NEW RECORD on 1916-05-18 - Mike Gibbons (Welterweight) @ 1868
NEW RECORD on 1916-09-30 - Les Darcy (Middleweight) @ 1889
NEW RECORD on 1917-06-06 - Ted 'Kid' Lewis (Welterweight) @ 1991
NEW RECORD on 1917-06-25 - Ted 'Kid' Lewis (Welterweight) @ 2000
NEW RECORD on 1919-01-27 - Harry Greb () @ 2355
NEW RECORD on 1919-01-31 - Harry Greb () @ 2413
NEW RECORD on 1919-06-23 - Harry Greb () @ 2500
NEW RECORD on 1919-09-01 - Harry Greb () @ 2520
NEW RECORD on 1919-01-27 - Harry Greb () @ 2356
NEW RECORD on 1919-01-31 - Harry Greb () @ 2414
NEW RECORD on 1919-06-23 - Harry Greb () @ 2609
NEW RECORD on 1919-09-01 - Harry Greb () @ 2677
NEW RECORD on 1922-05-23 - Harry Greb (Light Heavyweight) @ 3397
NEW RECORD on 1922-07-10 - Harry Greb () @ 3548
NEW RECORD on 1925-03-27 - Gene Tunney (Light Heavyweight) @ 4059
NEW RECORD on 1925-06-05 - Gene Tunney () @ 4326
NEW RECORD on 1934-06-14 - Max Baer (Heavyweight) @ 4696
NEW RECORD on 1938-06-22 - Joe Louis (Heavyweight) @ 5236
NEW RECORD on 1939-01-25 - Joe Louis (Heavyweight) @ 5969
NEW RECORD on 1939-06-28 - Joe Louis (Heavyweight) @ 6056
NEW RECORD on 1939-09-20 - Joe Louis (Heavyweight) @ 6220
NEW RECORD on 1941-06-18 - Joe Louis (Heavyweight) @ 6337
NEW RECORD on 1941-09-29 - Joe Louis (Heavyweight) @ 6562
NEW RECORD on 1955-06-22 - Archie Moore () @ 7206
NEW RECORD on 1955-09-21 - Rocky Marciano (Heavyweight) @ 8994
Posted: 09 Aug 2007, 11:17
by m1kee50
JCS What does the system call the biggest upset (i.e lowest rating beating highest rating)? please mate?
Alos if possible, what is the best win streak out there in terms of fights udefeated combined with opponent quality?
Posted: 09 Aug 2007, 12:07
by JCS
~ MIKEE ~ wrote:JCS What does the system call the biggest upset (i.e lowest rating beating highest rating)? please mate?
Alos if possible, what is the best win streak out there in terms of fights udefeated combined with opponent quality?
The opposition quality penalty is pretty steep in this system. Far too steep in my opinion, so I'm not sure how much that'd be worth.
However, I'd almost guarantee that the answer would be Marciano or Monzon.
I will post the biggest upset here soon. Points issue again. Guys in the past racked up huge points because opposition quality was better, so the biggest difference between winner and loser is likely going to be 50 years ago.
Its processing now, but the winner so far is
Code: Select all
1940-05-10 - Lew Jenkins (1031.5) - TKO - Lou Ambers (5297.3) @ [4265.8] Lightweight
