harry_rag Posted July 1, 2021 Share Posted July 1, 2021 12 minutes ago, waynecoyne said: presumably @harry_ragthe 50 factors in that more goals are scored in the second half (i'm assuming this is so) Well, it is true that most goals are scored in the second half so I suppose that factors in. The 50 isn't an assumed figure, it just is the average (or has been traditionally). If you look at a big enough sample of games then the average time of all the goals scored will be very close to 50. As I said, a game with 2.5 goals expected will have a goal minutes expectation of around 125. On that basis we can think of our 3 players at 5, 10 and 20 as having goal expectations of 0.1, 0.2 and 0.4. Those are the numbers I use for the purpose of getting equivalent anytime odds using poisson. waynecoyne 1 Quote Link to comment Share on other sites More sharing options...
harry_rag Posted July 1, 2021 Share Posted July 1, 2021 And so to the use of poisson! Basically if you have an average or an expectation for something you can use it to give you a probability for any given number of occurrences. Goals in football broadly follow the distribution but it's not ideal for forecasting match results because it underestimates the chances of the draw. It's less flawed for goals, be it match, team or player goals. For instance, it would equate that 2.5 goals game to odds for >2.5 goals of 2.19. In terms of our 3 players it would give the following odds for them scoring 1 or more goals in a game: 5 = 10.51, 10 = 5.52 and 20 = 3.03. So, in all cases bigger odds than our simple "rule of thumb" metrics because it takes account of the increasing chance that the player could score 2 or more goals. A player who scores 20 goals in a season is likely to have scored >1 in some of those games so will have scored in less than 20 of their games. I've just quickly applied poisson to Harry Kane last season (23 goals in 35 EPL games). It suggests he would have scored one or more goals in 17 games when it was actually 18, Exactly one goal would be expected 12 times (actually 13), exactly 2 goals 4 times (5) and 1 hat trick (0). Only one example but fairly typical of how it works. waynecoyne 1 Quote Link to comment Share on other sites More sharing options...
waynecoyne Posted July 1, 2021 Share Posted July 1, 2021 8 minutes ago, harry_rag said: And so to the use of poisson! Basically if you have an average or an expectation for something you can use it to give you a probability for any given number of occurrences. Goals in football broadly follow the distribution but it's not ideal for forecasting match results because it underestimates the chances of the draw. It's less flawed for goals, be it match, team or player goals. For instance, it would equate that 2.5 goals game to odds for >2.5 goals of 2.19. In terms of our 3 players it would give the following odds for them scoring 1 or more goals in a game: 5 = 10.51, 10 = 5.52 and 20 = 3.03. So, in all cases bigger odds than our simple "rule of thumb" metrics because it takes account of the increasing chance that the player could score 2 or more goals. A player who scores 20 goals in a season is likely to have scored >1 in some of those games so will have scored in less than 20 of their games. I've just quickly applied poisson to Harry Kane last season (23 goals in 35 EPL games). It suggests he would have scored one or more goals in 17 games when it was actually 18, Exactly one goal would be expected 12 times (actually 13), exactly 2 goals 4 times (5) and 1 hat trick (0). Only one example but fairly typical of how it works. so that equates to an extra 10% roughly? Quote Link to comment Share on other sites More sharing options...
harry_rag Posted July 1, 2021 Share Posted July 1, 2021 40 minutes ago, waynecoyne said: so that equates to an extra 10% roughly? Ha, it seems to though that may be random! The % seems to get smaller the higher the goal minutes price goes. You can replicate that to an extent by dividing by 1.8 instead of 2 in your equation, that gets you closer to the poisson price. I'm now playing about with your equation to see if it can be fine tuned further! waynecoyne 1 Quote Link to comment Share on other sites More sharing options...
harry_rag Posted July 1, 2021 Share Posted July 1, 2021 Ok, forget that 1.8 notion, use your original calculation but reduce the spread price by 10%, so instead of 90 - 7 it's 90 - 6.3. That gets you pretty close to the possion price it seems! The bigger the number the less accurate it becomes but it's pretty good for anything below 30. Having said that, if you've got Excel just use the poisson function and get a more accurate guide for any given price! 2 ways to use these comparisons; which is best value for a player you intend to back, buying the goal minutes or backing the player anytime or use the spreads to identify possible value in fixed odds prices, e.g. a player is 22-25 with one firm and 23-26 with another. The midpoint of the best prices is 24 so you take that as his true value and use it to determine a fair fixed odds price (2.62). That's what I use to do until I realised that the spread prices are pitched higher than they ought to be because most people prefer to buy. Assuming the prices are skewed for buyers, I'd take the lowest sell price rather than the midpoint, so 22 equates to 2.81. Anything bigger than 2.81 MIGHT be worth a second look. To be honest, 22 is probably still too high a number to be using to call it the true value, but it's a starting point when looking for potential value in the anytime market. There endeth today's lecture! Quote Link to comment Share on other sites More sharing options...
waynecoyne Posted July 1, 2021 Share Posted July 1, 2021 10 minutes ago, harry_rag said: Ok, forget that 1.8 notion, use your original calculation but reduce the spread price by 10%, so instead of 90 - 7 it's 90 - 6.3. That gets you pretty close to the possion price it seems! The bigger the number the less accurate it becomes but it's pretty good for anything below 30. Having said that, if you've got Excel just use the poisson function and get a more accurate guide for any given price! 2 ways to use these comparisons; which is best value for a player you intend to back, buying the goal minutes or backing the player anytime or use the spreads to identify possible value in fixed odds prices, e.g. a player is 22-25 with one firm and 23-26 with another. The midpoint of the best prices is 24 so you take that as his true value and use it to determine a fair fixed odds price (2.62). That's what I use to do until I realised that the spread prices are pitched higher than they ought to be because most people prefer to buy. Assuming the prices are skewed for buyers, I'd take the lowest sell price rather than the midpoint, so 22 equates to 2.81. Anything bigger than 2.81 MIGHT be worth a second look. To be honest, 22 is probably still too high a number to be using to call it the true value, but it's a starting point when looking for potential value in the anytime market. There endeth today's lecture! thanks for your efforts @harry_rag i did maths with stats at a level but that was 46 years ago? harry_rag 1 Quote Link to comment Share on other sites More sharing options...
harry_rag Posted July 1, 2021 Share Posted July 1, 2021 4 minutes ago, waynecoyne said: thanks for your efforts @harry_rag i did maths with stats at a level but that was 46 years ago? Probably trumps my slightly less old grade C O level at Maths! waynecoyne 1 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.