Boxing Computer Simulator Picks Juan Manuel Marquez

[Chups]

Marvin Langer's Super Computer Returns!
Boxing Computer Simulator Picks Juan Manuel Marquez
Marven Langer


Over the years I have developed my own home brewed computer simulator which predicts the outcomes of fights. Not to bore you with the details since the simulator is powered by some hefty statistical computations, bouts are simulated based on the following empirical and assumed data:
Each fighters tendencies to fight on the inside or outside.
Each Fighters' aggressiveness and tendency to foul
Punching power
Defense
Ability to recover from punishment
Tendency to cut.
Intangibles
How do I rate fighters you might ask. By watching hours of hours of film and discussing the fighters with other fight fans and experts.
This program is not a video game and has no visual luxuries. Again, it is purely a statistical processor which predicts the outcome of fights.
After running the characteristics of both Juan Manuel Marquez and Manny Pacquiao 100,000 times in anticipation of their March 15 rematch the results are as follow:
Juan Manuel Marquez Winner 55%
Manny Pacquiao Winner 40%
Draw 5%
These numbers are odd in the sense that a five percentage draw result is rather high. In standard simulations, the draw average runs around 1 or 2 percent. It might be a direct reflection of the even level of abilities between the two men.
Even more odd is the 55% win mark by Marquez with the draw percentage so high. This might be a reflection of many experts notion that Pacquiao's only chance to win is by knockout.
Further breakdowns:
Marquez Winner By KO 10%
Marquez Winner By Decision 90%
The computer, like experts,does not see Marquez knocking out the iron chinned Pacquiao.
Pacquiao Winner By KO 70%
Pacquiao Winner By Decision 30%
Again the computer does not see Pacquiao able to take a decision from Marquez.
During the last twelve months my simulation package has been predicting the winner of fights at an accuracy of 95%.

[Zerwas1]

My simulation in my head has an 95% chance of picking a winner in a fight, if I know the fighters.

[Jim Jeffries]

Didn't a boxing simulator pick Marciano to beat Ali?

Nuff said.

This fights still a pickem in my book.

[Motofan]

The simulator is right. Pac cant win without a knock out and JMM aint stopping Pac any time soon. JMM by SD.

[Gareth Ivanovic]

I hope the simulator is right, but it's such an even matchup it's hard to pick a winner. I agree Marquez by SD in a close fight.

[grayfist]

The assumptions can be flawed, thus, GIGO.

Quantification of punching power based on career-long KO rates can be very deceptive because it may not factor into the equation the quality of the KO victims. For example, Marquez has never stopped anyone that comes near the likes of Morales, and even himself. Morales was stopped twice by Pacquiao and stopping a Morales cannot be statistically similar to stopping, say, a Terdsak. Similarly, three KD's inflicted on someone like a JMM cannot be counted as of equal value to three KD's of anyone other than say a Barrera, a Morales, or some fighter of that level.

Pacquiao's capacity to recover may not have also been faithfully represented if such capacity is measured in terms of his inability to continue from the two stoppages early in his career. The circumstances are not going to be the same.

The assumptions also include opinions of experts. These subjective inputs may only have severely influenced the results; they may have even, perhaps, nullified the mathematics on punch rates and accuracy (if these numbers, which are a tad more reliable than other stats,e.g., so-called "tendencies to fight on the inside" and the unspecified "intangibles", have at all been factored, which I suspect they were not because no mention was made of them).

I think statistics on punch rates and accuracy are more helpful to simulation than opinions can be.

To me, opinions should be charted separately from any attempt at predicting results through an evaluation of stats derived from past performances. Putting them together makes the conclusion, at the very least, suspect.

Anyhow, without the aid of any simulator, I am guessing that the fight is much more a pick 'em than that simulator presents it to be.

[RodBarker]

The assumptions can be flawed, thus, GIGO.

Quantification of punching power based on career-long KO rates can be very deceptive because it may not factor into the equation the quality of the KO victims. For example, Marquez has never stopped anyone that comes near the likes of Morales, and even himself. Morales was stopped twice by Pacquiao and stopping a Morales cannot be statistically similar to stopping, say, a Terdsak. Similarly, three KD's inflicted on someone like a JMM cannot be counted as of equal value to three KD's of anyone other than say a Barrera, a Morales, or some fighter of that level.

Pacquiao's capacity to recover may not have also been faithfully represented if such capacity is measured in terms of his inability to continue from the two stoppages early in his career. The circumstances are not going to be the same.

The assumptions also include opinions of experts. These subjective inputs may only have severely influenced the results; they may have even, perhaps, nullified the mathematics on punch rates and accuracy (if these numbers, which are a tad more reliable than other stats,e.g., so-called "tendencies to fight on the inside" and the unspecified "intangibles", have at all been factored, which I suspect they were not because no mention was made of them).

I think statistics on punch rates and accuracy are more helpful to simulation than opinions can be.

To me, opinions should be charted separately from any attempt at predicting results through an evaluation of stats derived from past performances. Putting them together makes the conclusion, at the very least, suspect.

Anyhow, without the aid of any simulator, I am guessing that the fight is much more a pick 'em than that simulator presents it to be.

Interesting stuff ,, I programed a punch counter a while back and added into the equation the five criterias of judging with a 1 to 5 score on each of the criterias for each fighter round by round , and split the landed punches into just two for ease of use , left and right with a power punch button ,,,, it was very accurate and did not agree with judges at times , I might pick it up again and do some more work on it ,,, interested in what you are doing Grayfist .

[Dempsey Roll]

Your simulator sounds very similar to the inner workings of Title Bout Championship Boxing

[ippo]

^^ it probably is i have the same game.

[grayfist]

Interesting stuff ,, I programed a punch counter a while back and added into the equation the five criterias of judging with a 1 to 5 score on each of the criterias for each fighter round by round , and split the landed punches into just two for ease of use , left and right with a power punch button ,,,, it was very accurate and did not agree with judges at times , I might pick it up again and do some more work on it ,,, interested in what you are doing Grayfist .

Hi there Rod, ol' pal! Thanks!

I surely am interested to know how your project pans. Care to share, buddy?

Thanks again!