Testing The Accuracy and Predictions of The Computer

Collins2000 · Post by **Collins2000** » 16 Oct 2006, 23:22

IrishRufusMurphy wrote:I'm trying to compile a list of fights, at least eight in number, only have six listed up so far:

Shannon Briggs vs. Sergei Liakhovich

Floyd Mayweather vs. Carlos Baldomir

Wladimir Klitschko vs. Calvin Brock

John Ruiz vs. Ruslan Chagaev

Jermain Taylor vs. Kassim Ouma

Oleg Maskaev vs. Peter Okhello

Anyone else know of any fights coming up soon? Preferably big fights?

Fight schedule here

http://www.maxboxing.com/News/schedule.asp

HomicideHenry · Post by **HomicideHenry** » 16 Oct 2006, 23:36

Okies, I found a few more fights, let's make it ten simulations.

These will be the fights that will be simulated, and each simulation will have its own thread so people can make predictions and when fight night approaches, you will see if the computer was right or wrong.

Shannon Briggs vs. Sergei Liakhovich

Floyd Mayweather vs. Carlos Baldomir

Wladimir Klitschko vs. Calvin Brock

John Ruiz vs. Ruslan Chagaev

Jermain Taylor vs. Kassim Ouma

Oleg Maskaev vs. Peter Okhello

Matt Skelton vs. Audley Harrison

Chris Byrd vs. Eliseo Castillo

Evander Holyfield vs. Fres Oquendo

Brian Minto vs. Axel Schulz

All these fights will happen in the next two-three months. So it will be a great time table. Eight of the fights to be simulated are at HW and the other two are at MW and WW.

So, this thread is officially done with. The next few months will have new simulations with their own threads. And finally settle the issue of how accurate this simulator is, so that the HW tournament that was done before can finally have credibility and continue forward.

Peace!

cubedrum · Post by **cubedrum** » 04 Nov 2006, 17:39

Here we go. Lots of fights tonight.
As usual simulation based on 1,000 fights converted to %

Floyd Mayweather Jr vs Carlos Baldomir
Both men always in top condition

Mayweather decision 35.5%
Mayweather KO 58.7%
Baldomir decision .8%
Baldomir KO 3.1%
Draw 1.9%

The most clear cut pick in a major fight yet. PBF by KO.

Robert Guerrero vs Orlando Salido
Both men always in condition
Not sure but I think this will be a 10 round fight.

Guerrero decision 21.6%
Guerrero KO 9.0%
Salido decision 43.1
Salido KO 16.6%
Draw 9.7%

Salido takes a close, perhaps split decision.

Paul Williams vs Mauro Lucero
Both men in top condition

Williams decision 14.4%
Williams KO 82.4%
Lucero decision .4%
Lucero KO 2.4%
Draw .4%

Complete blowout for Williams. KO inside of 4.

Sergei Lyakhovich vs Shannon Briggs
Both men always in shape

Lyakhovich decision 21.2%
Lyakhovich KO 39.3%
Briggs decision 17.4
Briggs KO 19.1%
Draw 3.0%

The most likely result is a Lyakhovich KO, probably late. Briggs shapes up on paper better than I would have thought, I think because of Lyakhovich's below average mobility.

Juan Diaz vs Fernando Angulo
Both men in shape
Assuming this is a 10 round fight

Diaz decision 61.6%
Diaz KO 26.9%
Angulo decision 4.1%
Angulo KO 4.5%
Draw 2.9%

Diaz takes the unanimous decision.

Cory Spinks vs Rodney Jones
Spinks always in shape, Jones usually
Again I think this is going 10.

Spinks decision 49.8%
Spinks KO 8.3%
Jones decision 23.4%
Jones KO 8.2%
Draw 11.3%

Spinks moves up to JMW to face declining Jones. Jones' experience and natural JMW frame makes it a tough fight, but Spinks takes a unanimous decision.

Luis Collazo vs Hercules Kyvelos
Kyvelos known to break camp. Collazo in top condition

Collazo decision 65.3%
Collazo KO 25.2%
Kyvelos decision 3.5%
Kyvelos KO 3.4%
Draw 2.6%

Collazo takes the unanimous decision.

That's it for now. Tubbs and Akinwande's opponents tonight are too obscure to be rated.

Collins2000 · Post by **Collins2000** » 04 Nov 2006, 17:48

Sorry, mate. Irish has already done the prediction for Lyakhovich - Briggs and he claims over 1000 sims it predicts a win for Briggs.

Now you say it is picking Lyakhovich????

And you both claim you are running with the virgin settings, yet both predict different outcomes over 1000 sims.

One of you is telling porkies again.

PS, there is a specific thread for the 2 main bouts this weekend. It would be less confusing if you posted in those wouldn't it?

cubedrum · Post by **cubedrum** » 04 Nov 2006, 17:51

Irish probably didn't use the "inactive" training setting for Briggs. Hey Collins, try the demo mode for FREE for your self.

http://www.ootpdevelopments.com/tb/
http://ootpdevfiles.com/tb/tb2setup.exe

prometheus · Post by **prometheus** » 07 Nov 2006, 06:52

Hey cubedrum.
Looks as if the computer predictions did pretty well overall over the weekend

cubedrum · Post by **cubedrum** » 07 Nov 2006, 19:16

Well, I missed the conditions on the Briggs fight. He had more in the tank than I would have thought. My bad. Would have been a clean sweep otherwise. More to come.

prometheus · Post by **prometheus** » 07 Nov 2006, 21:33

cubedrum wrote:Well, I missed the conditions on the Briggs fight. He had more in the tank than I would have thought. My bad. Would have been a clean sweep otherwise. More to come.

Looking forward to seeing how these predictions perform over a larger sample, obviously no method of prediction is foolproof or perfect but it would be interesting to see if it can noticeably outperform the general publics 'predictions'

Collins2000 · Post by **Collins2000** » 07 Nov 2006, 21:43

prometheus wrote:
cubedrum wrote:Well, I missed the conditions on the Briggs fight. He had more in the tank than I would have thought. My bad. Would have been a clean sweep otherwise. More to come.
Looking forward to seeing how these predictions perform over a larger sample, obviously no method of prediction is foolproof or perfect but it would be interesting to see if it can noticeably outperform the general publics 'predictions'

It is infallible. Even though Cubey missed out on the Briggs fight, Irish_Ruper_Murdoch used the same simulator and he got it right!

cubedrum · Post by **cubedrum** » 07 Nov 2006, 22:16

Collins2000 wrote:
prometheus wrote:
cubedrum wrote:Well, I missed the conditions on the Briggs fight. He had more in the tank than I would have thought. My bad. Would have been a clean sweep otherwise. More to come.
Looking forward to seeing how these predictions perform over a larger sample, obviously no method of prediction is foolproof or perfect but it would be interesting to see if it can noticeably outperform the general publics 'predictions'
It is infallible. Even though Cubey missed out on the Briggs fight, Irish_Ruper_Murdoch used the same simulator and he got it right!

Well it's certainly not infallible. It is only as accurate as the boxer ratings are. But I must say that the ratings, produced by Jim Trunzo and Boxrec's own Dan Cuoco, are spot on perhaps 90% of the time.

Collins2000 · Post by **Collins2000** » 07 Nov 2006, 22:40

cubedrum wrote:
Collins2000 wrote:
prometheus wrote: Looking forward to seeing how these predictions perform over a larger sample, obviously no method of prediction is foolproof or perfect but it would be interesting to see if it can noticeably outperform the general publics 'predictions'
It is infallible. Even though Cubey missed out on the Briggs fight, Irish_Ruper_Murdoch used the same simulator and he got it right!

Well it's certainly not infallible. It is only as accurate as the boxer ratings are. But I must say that the ratings, produced by Jim Trunzo and Boxrec's own Dan Cuoco, are spot on perhaps 90% of the time.

Cubey, what I can't understand is why in some posts (this one for instance) you imply the simulator is a 'sealed unit' that works without user input, but in other posts (see above) you suggest user input is all important.

cubedrum · Post by **cubedrum** » 07 Nov 2006, 23:26

Here's the various categories of adjustments in the sim.

Career stage adjustments:
Beginning
Pre-Prime
Prime
Post-Prime
End

And Training adjustments:
Top condition
Overtrained
Undertrained
Broke training
Inactive
Aging and inconsistent
Badly overweight
Fights too frequently
Emotional problems
Lacks focus
Unprepared
Just needs a paycheck.

And if it applies you can designate a fighter as:
Hometown favorite
Defending champion

Also there are variables with various trainers and cutmen in the sport.

As you can see some of the adjustments are applied in a subjective manner, though they can be viable factors in predicting a fight. The engine is very versatile in this way. I urge you to check it out for yourself. :P

Collins2000 · Post by **Collins2000** » 08 Nov 2006, 04:27

cubedrum wrote:Here's the various categories of adjustments in the sim.

Career stage adjustments:
Beginning
Pre-Prime
Prime
Post-Prime
End

And Training adjustments:
Top condition
Overtrained
Undertrained
Broke training
Inactive
Aging and inconsistent
Badly overweight
Fights too frequently
Emotional problems
Lacks focus
Unprepared
Just needs a paycheck.

And if it applies you can designate a fighter as:
Hometown favorite
Defending champion

Also there are variables with various trainers and cutmen in the sport.

As you can see some of the adjustments are applied in a subjective manner, though they can be viable factors in predicting a fight. The engine is very versatile in this way. I urge you to check it out for yourself. :P

I'll check it out one day but for the time being I am just interested in seeing what you and Irish set the variables to. Up until now you've never actually said what they were, let alone what you set them to.

Now we are making some progress as to how the simulator can come up with two completely different outcomes even over 1000 runs.

Regardless of how good the engine is, it's now patently obvious that how you set the variables determines the outcome. I can't understand why I've had to continually cross-exam you guys to get this info.

cubedrum · Post by **cubedrum** » 08 Nov 2006, 14:53

Well the career stage is pre set on active fighters. It's more of a way to "age" fighters in a universe or as a way to reproduce certain historical effects (Louis vs Marciano for example. Louis in his prime wins about half the time).

The training effects, while sometimes profound, are used much less when the fights (most of them that I've simmed here anyway) are between prime world class fighters in top condition. I used it (and broke down which conditions I used BTW) on the McCall-Wilson AND Peter-Toney fight. McCall is known to be mentally unstable even in his prime, so to ignore that factor completely is to ignore pertinent data, which results in a less accurate computer sim. Same thing with Toney and his fat, lazy training habits. Not the only pertinent factor, but certainly an important one.

prometheus · Post by **prometheus** » 08 Nov 2006, 23:52

Hi again cubedrum
Any possibility of running some sims of the weekend's big fights and putting up the predictions.
I'm still interested in seeing how they pan out

cubedrum · Post by **cubedrum** » 09 Nov 2006, 03:28

You betcha. I'll post them on Friday morning.

prometheus · Post by **prometheus** » 09 Nov 2006, 21:30

cubedrum wrote:You betcha. I'll post them on Friday morning.

cubedrum · Post by **cubedrum** » 10 Nov 2006, 00:52

Here's this weekend's big fights as predicted by Titlebout 2.
As usual based on 1,000 bouts converted to %

Evander Holyfield vs Fres Oquendo
This one was tricky, because there are no current ratings for Holyfield.
I took his 2004 rating and adjusted it according to the stock formula from
Post-Prime to End.
Oquendo is rated at Post-Prime.
Oquendo in Top condition
Holyfield in Inactive condition (he's earned it)

Oquendo decision 60.4%
Oquendo KO 31.7%
Holyfield decision .9%
Holyfield KO 5.5%
Draw 1.5%

Oquendo by a unanimous and disturbing to watch decision.

Oscar Diaz vs Golden Johnson
Each has had poor camps but we'll call it top shape for both because either may come in undertrained. Another tricky one because Diaz moves up to welterweight. Johnson, who seems like he's been around forever is only 32.

Diaz decision 17.1%
Diaz KO 10.3%
Johnson decision 51.5%
Johnson KO 14.3%
Draw 6.8%

Johnson takes an upset decision. Diaz leaves his punch behind when he moves up to welterweight.

Wladimir Klitschko vs Calvin Brock
Hoo baby is this one close on paper! Both men known to usually be in top condition.

Klitschko decision 12.2%
Klitschko KO 40.0%
Brock decision 8.7%
Brock KO 37.1%
Draw 2.0%

This is virtually a pick 'em with a Klitschko KO barely being the most likely result. One thing's for sure the game doesn't think there will be a decision!

Kevin Kelley vs Manuel Medina

Kelley moves up to junior lightweight. Kelley is 39, Medina 35. I'm assuming this one is going 10 rounds.

Medina decision 44.5%
Medina KO 20.8%
Kelley decision 19.0%
Kelley KO 8.5%
Draw 7.2%

Medina, who has been a JLW for a long time is just too much for the faded Kelley who moves up to JLW. Medina takes a unanimous decision.

Laila Ali vs Shelly Burton

Ali, arguably the best female boxer of all time and a natural SMW, takes on tomato can Shelly Burton who usually fights a JMW.

Ali decision 35.2%
Ali KO 63.9%
Burton decision .4%
Burton KO .3%
Draw .2%

Huge mismatch favoring Ali. Look for an early KO.

No rating for Bernard Dunne who's fighting Esham Pickering.
No rating for Eliseo Castillo (not to be confused with Eliecer Castillo) who's fighting Chris Byrd.
No rating for Michael Hunter, who's fighting Steve Molitor.
No rating for Gabriel Elizondo who's fighting Jose Navarro.
These fights were listed at maxboxing. Opponents may change in real fights.

I'll get the weekday's fights on Monday morning. Thanks for your interest everyone!

Ezzard · Post by **Ezzard** » 10 Nov 2006, 06:37

cubedrum wrote:
Collins2000 wrote:
prometheus wrote: Looking forward to seeing how these predictions perform over a larger sample, obviously no method of prediction is foolproof or perfect but it would be interesting to see if it can noticeably outperform the general publics 'predictions'
It is infallible. Even though Cubey missed out on the Briggs fight, Irish_Ruper_Murdoch used the same simulator and he got it right!

Well it's certainly not infallible. It is only as accurate as the boxer ratings are. But I must say that the ratings, produced by Jim Trunzo and Boxrec's own Dan Cuoco, are spot on perhaps 90% of the time.

Cube

From the non-active historical fighters who do you think has been rated incorrctly?

cubedrum · Post by **cubedrum** » 10 Nov 2006, 12:54

Ezzard wrote:
cubedrum wrote:
Collins2000 wrote: It is infallible. Even though Cubey missed out on the Briggs fight, Irish_Ruper_Murdoch used the same simulator and he got it right!

Well it's certainly not infallible. It is only as accurate as the boxer ratings are. But I must say that the ratings, produced by Jim Trunzo and Boxrec's own Dan Cuoco, are spot on perhaps 90% of the time.
Cube

From the non-active historical fighters who do you think has been rated incorrctly?

I've found that some fighter's rating profiles are a little too effecient and some are too inefficient. None of them are that far off in my opinion. The fighters I'm going to list are just ones I've noticed through extensive playing of the game.

Overrated fighters I've noticed include:
HW Lem Franklin
HW Bob Cleroux
HW Jimmy Young
Mike McCallum at JMW
LHW Maxie Rosenbloom
LHW Tommy Loughran

And several old timers seem overrated.

Anyone with a great chin is a bit underrated (meaning that the chin takes up too large a portion of the final rating) including
Heavyweight George Chuvalo
Middleweight Jake LaMotta

These are off the top of my head. The ratings are on maybe 97% of the time, but I wanted a margin for error in my earlier statement.

prometheus · Post by **prometheus** » 11 Nov 2006, 21:34

cubedrum wrote:Here's this weekend's big fights as predicted by Titlebout 2.
As usual based on 1,000 bouts converted to %

Oscar Diaz vs Golden Johnson
Each has had poor camps but we'll call it top shape for both because either may come in undertrained. Another tricky one because Diaz moves up to welterweight. Johnson, who seems like he's been around forever is only 32.

Diaz decision 17.1%
Diaz KO 10.3%
Johnson decision 51.5%
Johnson KO 14.3%
Draw 6.8%

Johnson takes an upset decision. Diaz leaves his punch behind when he moves up to welterweight.

Upset very nicely called
Congrats

cubedrum · Post by **cubedrum** » 12 Nov 2006, 16:22

4 for 5 this weekend.

The computer's total record (with me at the controls) stands at 13-3 or 81.2% winners. If not for my own mistakes on picking the training conditions (tough when you don't see the weigh in and the fighter has a history of training poorly or is old and unrated) it could easily be 15-1 or 94%.

Some of the fights I ran didn't actually take place and the opponents changed in the 11th hour.

This weeks fights by Monday afternoon.

prometheus · Post by **prometheus** » 12 Nov 2006, 20:34

Now that was a very impressive display, congrats on the results, looking forward to seeing more predictions.

Great stuff 8)

HomicideHenry · Post by **HomicideHenry** » 14 Nov 2006, 17:37

AM kinda disappointed, the computer was wrong, my predictions anyways, with Holyfield winning a decision against Fres Oquendo. Guess that goes to show that computers, no matter how accurate, cant seem to determine the power of the human spirit.

prometheus · Post by **prometheus** » 14 Nov 2006, 21:38

IrishRufusMurphy wrote:AM kinda disappointed, the computer was wrong, my predictions anyways, with Holyfield winning a decision against Fres Oquendo. Guess that goes to show that computers, no matter how accurate, cant seem to determine the power of the human spirit.

I wouldn't be too disappointed, a single result is not overly meaningful when testing the effectiveness of predictive ability
If you compare cubedrum's sims for last weekends fights, you'll see that the Holy fight was the only one that the computer got 'wrong'.

Though, in all fairness, you'll see that by cube's sims, the computer gave Holy around a 6% chance of winning, small but still a chance.
Try a few more sims for the upcoming fights and see how it gets on, basically, nobody and nothing can predict the future with anything approaching 100%.

If it could, we'd all be millionaires very quickly and easily