BoxRec

You should really be pitching the computer's predictions against polls on the forum. The computer should be trying to beat human predictions.

HairyBoxe24 wrote:Stop bloody having a go at him, this isn't about personal battles.

THE COMPUTER CLEARLY PICKED PETER IF TONEY WAS OVER 230, END OF.
Nothing else matters about who said what, this is about the computer and it clearly picked PETER, now stop having a go about it.


Also i'd like to add although holyfield was a gimme, it even got the round right so tbh its 4-0.

No, the computer didn't clearly pick Peter if Toney was over 230. It does matter who said what and also importantly when.

IrishRufus Murphy kept saying before the fight that the computer was picking Toney. I saw that cubedrum was saying that if Toney was "undertrained" - that the computer was predicting that Peter would win. It was cubedrum's opinion that anything over 230 or so, would mean Toney was undertrained. The Computer didn't specify that 230 or more meant Toney undertrained and that under 230 Toney was in good shape.

I was concerned that it wasn't clear who the computer was picking, so I asked point blank at 2:48 am on Sept 2,who was the computer picking?
Just 4 minutes later, IrishRufusMurphy said that the computer was picking Toney. cubedrum didn't try to correct this until AFTER THE FIGHT.
Even at 4:40 am on September 3, after the fight, IrishRufusMurphy said "Samuel Peter makes the the upset winning a 12 round decision over James Toney, breaking the computers 2 fight win prediction streak.
Then later he started to change his story.
It's obvious that he is trying to have it both ways. If Toney would have won, he would have pointed out that he had said before the fight that the computer had picked Toney to win.

As for Holyfield-Bates, it's not that the fight was easy to predict that bothers me the most.
What bothers me (and others) is that IrishRufusMurphy told us the computer's "prediction" AFTER the Holyfield-Bates fight was over!
Of course it was right and had the round right.
That doesn't count. Anyone can "predict" who won a fight AFTER it happens!

The computer clearly predicted 2 out of 3 fights correctly so far.

I already said I take blame for saying the computer picked Toney. I have explained this time and time again, NOT ALL FIGHTS ARE GOING TO BE WITH PARTICIPANTS WHO ARE ALWAYS IN SHAPE. The computer wasn't wrong in picking Peter if Toney was under-trained, we all knew at 233 he was heavier than he should have been.

Besides I have said instead of 10 fights to run I would run 12 to make up for both Holyfield and Toney---mistakes I made. I have apologised over and over again for this. I will not do it again, no I will not apologise for something I already said I was sorry for.

And if the computer had picked just Toney and Peter won, I would have counted it for Peter, or any of these fights for that matter, irregardless if the computer was wrong or right I would put the computer's predictions based against the actual results, and if the computer is wrong IT WILL BE WRITTEN DOWN AS WRONG, if it is predicted right, then IT WILL BE WRITTEN DOWN AS CORRECT.

The Holyfield Gimme and the Toney-Peter fights have been THROWN OUT, thats why I am running 12 rather than 10 fights, because IT IS MY OWN FAULT for the errors.

Now let's give it a rest...please?

Whether you want to look at it as being 4-0, 2-1 or whatever this testing is still going on as planned. And as far as THE HUMAN PREDICTION vs THE COMPUTER PREDICTION...don't make me laugh because I have asked constantly for the 'experts' here at BoxRec to throw in their opinions on upcoming fights and nobody really at all stood up to give their two cents.

ANYWAYS...since Holyfield 'gimme' is out and so is the Peter-Toney fight [yes again it was my own fault for the flubs not the comp] the record for these predictions is now 2-0 [Maskaev-Rahman/McCall-Wilson] with 10 more fights to go.

if Cube/Ezzard or anyone would so kindly do the computer predictions for the fights I written earlier, it would be greately appreciated:

Shannon Briggs vs Sergei Liakhovich- WBO title
Floyd Mayweather vs Carlos Baldomir- Welterweight title
Marco Antonio Barrera vs Rocky Juarez
Manny Pacquiao vs Erik Morales
Kevin McBride vs Mike Mollo

Mind you ALOT of these fights are happening on THE 16th, three days away, so if your going to do it, do it quickly.

Your running these on Title Bout correct? I have it and will run the fights.

Yes we are using Title Bout the rankings Cube been using are of the 2005-2006 ratings, so I hope you do as well...if not I guess we can manage, considering the computer still had the foresight to predict the winner of Maskaev-Rahman though it was 2003 rakings.

Thanks you very much, I would GREATLY appreciate it, if these can be done today or early tommorow...else we'll miss the dead line for these fights. This tournament has had enough flubs as it is...

Here's tomorrow's fights.
(based on 1,000 fights converted to %)

Marco Antonio Barrera vs Rocky Juarez
Both men are known to train like monks.

Barrera KO: 33.3%
Barrera Dec: 37.1%
Juarez KO: 14.1%
Juarez Dec: 10.6%
Draw : 4.9%

Barrera comes into this one focused and takes a unanimous decision.


Jorge Barrios vs Joan Guzman
Again both men historically very serious about training

Barrios KO: 59.3%
Barrios Dec: 16.9%
Guzman KO: 15.5%
Guzman Dec: 6.1%
Draw: 2.2%

Guzman has trouble retaining his power at a new weight class for him. A surprisingly clear cut prediction. The computer picks Barrios to easily retain his WBO JLW title by knockout.


Jhonny Gonzales vs Israel Vasquez
Both always in condition

Gonzales KO: 13.1%
Gonzales Dec: 6.3%
Vasquez KO: 64.2%
Vazquez Dec: 14.7%
Draw: 1.7%

Another case of Gonzales fighting at a new weight for him, and not carrying his power with him. Clear cut knockout victory for Israel Vasquez.


I listed the results a little different this time because the McCall-Wilson results were read by me a little wrong. This way is better.

Okies...I guess we will have to wait and see who wins these fights, though two of the three predicted are fights I didn't ask for lol

I thought for sure the Pacquia/Morales fight was on today...or maybe I'm mistaken...but o well good job Cube.

IrishRufusMurphy wrote:Okies...I guess we will have to wait and see who wins these fights, though two of the three predicted are fights I didn't ask for lol
I thought for sure the Pacquia/Morales fight was on today...or maybe I'm mistaken...but o well good job Cube.

That's OK, Rupert. If the computer gets them right I'm sure you will count them and if the computer gets any wrong just call it a no contest like before.

Collins....if they're right, they are right...they weren't really the fights I asked for, but we do need more fights listed...AND...if they are wrong, i will count them as being wrong

So kiss my ass Collins.

Vazquez beats Gonzales....

Guzman beats Barrios...

that's 1-1 for the computer...the computers first clear loss in this tournament.

and....


Barrera beats Juarez....

thats 2-1 for the computer in these 3 fights....making the computer...4-1 total [excluding the Toney/Holyfield] fights. 8)

Shannon Briggs vs Sergei Liakhovich- WBO title
Floyd Mayweather vs Carlos Baldomir- Welterweight title
Manny Pacquiao vs Erik Morales
Kevin McBride vs Mike Mollo


^^^^^^^^^any of these would be the NEXT fights that I would like
to be seen simulated on the Title Bout game. Also, the Nicolay valuev-Monte Barrett fight was simulated some posts ago, so that would make an even NINE fights, with three more to simulate to make the 12. 8)

at 75% accuracy it's running at about the rate of the average educated human mind. But time will tell...we need many many trials to establish parameters.

I agree...more tests need to be done and all...going to run 12 fights total, we need nine more fights to simulate to finish the testing...

I am proud thus far how the computer is going...am disapointed that it didn't predict one of the fighst last night correctly, but computers ain't full-proof and I think what a computer can't measure is a fighters heart...and I think in that case last night, that was the variable.

The goal in mind was to have the computer run at 75% accuracy, so far it has been running at that rate...but the goal is still 12 simulations, so if it is still 75% or higher then the goal has been achieved.

Sorry to 'triple post' but if Cube or Ezzard or anyone else with TITLE BOUT and with recent rankings can do the fights I listed to be simulated, or any fights in particular to their fancy [that are upcoming bouts] I would greatly appreciate it.

It seems that Wladimir Klitschko [who side stepped Shannon Briggs] will now take on undefeated contender Calvin Brock. 8) This should be another fight to be simulated by the computer.

sorry that i havent kept up with this...but Nicolay Valuev knocked out Monte Barrett in the 11th round...the computer picked Valuev to win, as cube had done the simulation for that particular fight a long while ago.

So...another win for the computer.

Note- The computer has been running at 75% accuracy, which was the original goal in this project.

Computer nailed the Cazares fight and the Valuev fight. The computer is now 6-1. I didn't get a chance to run Corrales-Casamayor, but I'm sure the computer would say it was a pick-em. I could still run it if no one believes I would doctor the results which I wouldn't.

Cube

No I havent been posting as much...havent really came on here as much either...Cube, I reckon we have 5 more fights to simulate...if you would be kind enough to do so, then this testing would be completed.

Be interesting to put up some of the computers ratings of fighters. Of course it would have to be translated for a lay person. Some of the ratings would be farily easy to put across, like punch power for example. The ratings would make interesting debating points.

I'll have a look and see what I can muster up.

I think whats interesting, as this was one of the goals, was to see how accurate this simulator was to BoxRec's computer rankings. After 7 simulations, we've only had one miss. With 5 more simulations [12] we could then see how accurate TITLE BOUT is, and see where BoxRec needs to change.

Mind you, if the TITLE BOUT simulator proves to be accurate [the goal was 75% accuracy], then BoxRec's rankings would have to change. As its top two HW's are Joe Louis and Muhammad Ali. As in my HW tournament, the final two were Ali and Marciano, with Joe Louis being third.

And of course, after testing is done, run the final fight [Marciano/Ali] and see who is the ATG HW CHAMPION and then do the other division tournaments. BUT, only if the simulator is accurate enough.

IrishRufusMurphy wrote:I think whats interesting, as this was one of the goals, was to see how accurate this simulator was to BoxRec's computer rankings. After 7 simulations, we've only had one miss. With 5 more simulations [12] we could then see how accurate TITLE BOUT is, and see where BoxRec needs to change.

Mind you, if the TITLE BOUT simulator proves to be accurate [the goal was 75% accuracy], then BoxRec's rankings would have to change. As its top two HW's are Joe Louis and Muhammad Ali. As in my HW tournament, the final two were Ali and Marciano, with Joe Louis being third.

And of course, after testing is done, run the final fight [Marciano/Ali] and see who is the ATG HW CHAMPION and then do the other division tournaments. BUT, only if the simulator is accurate enough.

Hang on, just because the compuer program gets some contemporary fights right it doesn't mean that the classic section fighters match up. the game engine is excellent but its how well the fighters are rated that counts. Title Bout does its best to put everyone on a level because that's the fun. Some people here will tell you that putting fighters from vastly different eras together is a hazardous thing to do. A lot of people think that lennox Lewis would decimate Fitzsimmons, others don't, of course, but there is an issue here.

Also Title Bouts top 2 HWs are Ali and Louis. The computer program takes into account all possibilities. If Ali fought Marciano a million times then Rocky would win some by KO. If you run the fight once and happen to get Marciano WKO1 this is not the definitive result by any means (it really could eb that 1 in a million result). Anything is possible in boxing and Title Bout does its best to model this.

Off the top of my head there is an overall score for boxers that is calculated by taking into consideration all factors and gives users a sense of who is at a similar level when looking at match ups. The highest mark is 15. As I remember, at HW, Ali is 15/15; Louis 14/15; Johnson/Dempsey/Marciano/Tunney 13/15; Holmes/Liston/Foreman/Frazier/Holyfield/Tyson 12/15.

Think of it this way, if Marciano beats Ali is it because he was a better fighter or because he is overrated in the game, or because Ali is underrated???

I think the only true way to determine how accurate the simulators are to predict upcoming fights. It's harder, in my opinion, to guess who would win an upcoming fight, than a 'fantasy fight' only on the condition that with past fighters you can review their records and see when they were their best and worst as a fighter---with fighters who are still active, you can't be for sure who is better because they are incomplete.

Take for example the first SUPERFIGHT in 1969. Ali was never really finished as a fighter, and alot of things [how great his chin was etc] were in question of his abilities, while Marciano hadn't fought in nearly 15 years, his record and fights could be reviewed and his strengths and weaknesses and prime years were factors that were known.

As a result the computer picked Marciano to win via KO. Now Ali has been retired since 1981, so everything that was in question is answered. Now we can review both mens strengths and weaknesses.

Now with predicting upcoming fights...there are questions still left unanswered because the fighters careers are not yet over with. I figured if the computer is accurate enough to predict fights with guys whose careers are still unanswered, then the computer surely should be able to predict the outcome of fighters who are easier to look over.

Now...one of the goals I have with this testing is to see how 'accurate' not only my findings are, but how accurate BoxRec's computer rankings are.

My top three fighters in my HW tournament were Ali, Marciano and Louis. The top five HW's on BoxRec's rankings are Ali, Tunney, Marciano, Johnson and Louis.

The only argument I make is that BoxRec's rankings take too much into count [in the case of Tunney] of his LHW career, and due to that he gets all these points amassed and then throw in his brief HW career. This computer system that BoxRec uses is based on points.

Mine is based on probability, 129 variables and of course the records of the fighters, plus their rankings in the eras they were apart of. I think also that my tournament is much more accurate than Murry Woroner's, but then again Ali's career was not complete, and had it been, it would have been ALI vs MARCIANO in the finals and not DEMPSEY vs MARCIANO.

Not disrespect for Jack Dempsey, but I think my tournaments final three of Ali, Marciano and Louis are more accurate than Woroner's that was Marciano, Dempsey and Jefferies.

Anyways, sorry I had to get that off my chest. I would like to see the computer determine the final five fights to make this experiment complete. Speaking of which, here are a few to list off, to take into consideration:

Shannon Briggs vs Sergei Liakhovich-WBO HW title

Floyd Mayweather vs Carlos Baldomir- IBF WW title

Jermain Taylor vs Kassim Ouma- MW title

Markus Beyer vs Mikkel Kessler- WBC/WBA SMW title(s) October 14th

Wladimir Klitschko vs Calvin Brock- IBF HW title