How to configure a ML

[a4u2fear]

Team has an estimated win pct of 60% vs "any" competition
Opponent has an estimated win pct of 50% vs "any" competition

what would be the proper ML for this game ?

[Waterstpub87]

-150/+150. Assuming the average "any" competition is 50%

[gojetsgomoxies]

because the opponent is 50% WP (average or any team), you can just use the 60% WP for the team as the expected WP......

what is more interesting is what is the expected winning % of a 70% WP team playing a 35% WP team. not sure there is a right answer. in baseball, the answer is probably 83% or something like that. but i can see situations in local high school football where the WP = 100% (often something like high school football extreme tiering, A kills B kills C kills D etc to team H.. so that would be 8 team league where almost all games are deterministic. you could relax that assumption and have fewer tiers but you could certainly have HS footbal situation where the 70% team is 100% to beat the 35% team.

[Cookie Monster]

Bill James invented the log5 formula many years ago. Check it: https://en.wikipedia.org/wiki/Log5

[lleytian3]

the following calculator will give you the proper ML for a 50% or 60% win pct. just type the win pct in the implied probability and you'll get the ML

https://www.sportsbookreview.com/picks/tools/odds-converter/

[turbobets]

Is it proper to use the probability of winning an event that has multiple participants to formulate head to head odds?

Probability to win against all participants:

Participant A = 20%
Participant B = 50%

Head to head:

Participant A = 20%/(20%+50%) = 28.6%
Participant B = 50%/(20%+50%) = 71.4%

Am I on the right track?

[Bsims]

Is it proper to use the probability of winning an event that has multiple participants to formulate head to head odds?

Probability to win against all participants:

Participant A = 20%
Participant B = 50%

Head to head:

Participant A = 20%/(20%+50%) = 28.6%
Participant B = 50%/(20%+50%) = 71.4%

Am I on the right track?

That is a simple approach that I have used in the past. Another alternative is to use Bill James log 5 approach (A-AB)/(A+B-2AB). The results would be Participant A = 20%, Participant B = 80%.

[turbobets]

Thanks for reply Bsims. I was thinking this would be useful when looking for edges on individual sports like motor racing, golf and tennis where match up odds are offered.

[Bsims]

Thanks for reply Bsims. I was thinking this would be useful when looking for edges on individual sports like motor racing, golf and tennis where match up odds are offered.

On the surface they might be useful. But probably not. I look at participants (teams or individuals) in events as likely to perform at potentially different levels. This varies from above average to average to below average. One could conceivably consider plotting their potential performance level on somewhat of a normal distribution curve. Now the odds of winning an event would have to inherently imply that the participant would perform at their top level. (One would assume that someone in a multi participant event would perform at peak level.) When considering individual matchups, one should consider comparing the participants at varying levels of performances. Hence different odds relationship.

[lleytian3]

On the surface they might be useful. But probably not. I look at participants (teams or individuals) in events as likely to perform at potentially different levels. This varies from above average to average to below average. One could conceivably consider plotting their potential performance level on somewhat of a normal distribution curve. Now the odds of winning an event would have to inherently imply that the participant would perform at their top level. (One would assume that someone in a multi participant event would perform at peak level.) When considering individual matchups, one should consider comparing the participants at varying levels of performances. Hence different odds relationship.

This is great information.

So for tennis specifically, if I can estimate that player A odds on their worst day would be better than player B on their best day, is there value to bet that if I can find the specific odds that fits this hypothetical?

[semibluff]

Is it proper to use the probability of winning an event that has multiple participants to formulate head to head odds?

Probability to win against all participants:

Participant A = 20%
Participant B = 50%

Head to head:

Participant A = 20%/(20%+50%) = 28.6%
Participant B = 50%/(20%+50%) = 71.4%

Am I on the right track?

I'd say it was dangerously wrong to take that approach and you would likely create some horribly wrong lines if you did. Leicester City might be 200/1 to win the English Premier League. West Ham Utd might be 2000/1 to win the title. Taking those lines and thinking Leicester would be 10 X better than West Ham or 10 X more likely to win a game against them would be extremely unsound. There will be times during the season when 1 team is firing on all cylinders whilst the other couldn't beat a rug with a stick. There will be other times when the situation will be completely reversed.

Consider this: when you handicap an outcome you'll factor everything in and come to a probability determination. That determination will probably be wrong, but hopefully by only a small margin. If you handicap 1 overall value setting for a team you will make 1 error. If you make the same determination for another team you now have a 2nd error. If you try to create a formula that will correlate a probable outcome using 2 bits of data you will create a 3rd error. It's possible the 3 errors will even themselves out but statistically there's a 1 in 8 chance of all 3 errors favouring 1 team and a 1 in 8 chance they'll all favour the other team. The problem is that it is these extremely favourable outcomes that will be most enticing to bet since you appear to be beating the market by a wide margin.

I guess there must be some way of doing it or there wouldn't be successful volume bettors. That said my advice is still to develop your own handicapping skills rather than trusting to formulae. Math is there to speed up the process and make things easier. It's not a substitute for knowledge and judgement.

[lleytian3]

I'd say it was dangerously wrong to take that approach and you would likely create some horribly wrong lines if you did. Leicester City might be 200/1 to win the English Premier League. West Ham Utd might be 2000/1 to win the title. Taking those lines and thinking Leicester would be 10 X better than West Ham or 10 X more likely to win a game against them would be extremely unsound. There will be times during the season when 1 team is firing on all cylinders whilst the other couldn't beat a rug with a stick. There will be other times when the situation will be completely reversed.

Consider this: when you handicap an outcome you'll factor everything in and come to a probability determination. That determination will probably be wrong, but hopefully by only a small margin. If you handicap 1 overall value setting for a team you will make 1 error. If you make the same determination for another team you now have a 2nd error. If you try to create a formula that will correlate a probable outcome using 2 bits of data you will create a 3rd error. It's possible the 3 errors will even themselves out but statistically there's a 1 in 8 chance of all 3 errors favouring 1 team and a 1 in 8 chance they'll all favour the other team. The problem is that it is these extremely favourable outcomes that will be most enticing to bet since you appear to be beating the market by a wide margin.

I guess there must be some way of doing it or there wouldn't be successful volume bettors. That said my advice is still to develop your own handicapping skills rather than trusting to formulae. Math is there to speed up the process and make things easier. It's not a substitute for knowledge and judgement.

I agree with the last part about not using math to substitute for knowledge.

But I'm a little confused on your handicapping theory and making errors. So when you handicap a typical 1 on 1 matchup you have numerous factors that goes into your handicap. Or you should have a lot of different factors. So are you saying each factor is potentially its own separate error? Or are you saying after you complete your handicap on a specific team, that is one potential error?

[turbobets]

Definitely would not use season long futures odds to formulate head to head match odds. I was considering odds to win an event to either formulate or compare head to head match odds in that same event.

[turbobets]

I experimented with horse racing match ups. I used TVG's odds to win and plugged them into A/(A+B).

So far...

Total Races Wagered = 170
Predicted Win Pct = 64.6%
Actual Win Pct = 62.9%