Understanding Variance in Results

[Cards37]

Hi,

I'm trying to build a few simple calculations into my tracking system to highlight variance to help shed some light on just how much impact it is having. I want to reinforce some first principles to make sure I'm not fundamentally misunderstanding.

So if a favourite is at $1.50, it has a 66.67% chance to win. If it wins, it has actually performed 0.33 above expectation. If it loses, its 0.67 below. Depending on the actual outcome, record either of those results. If the book is 100%, in the long run across a large number of events, the sum of all these should be 0. If the total is positive, favourites are winning above expectation. If its negative, favourites are losing more than expectation. Is this correct?

I have a small sample size so far of about 50 events, the running total is 3.714, which says favourites are running well above expectation.

I might be making a fundamental mistake here so would appreciate advice, thanks!

[Atho55]

You can do it by comparing how many you would expect to win at each BSP v how many actually won at the same BSP.

Here is an example. AUS races excl Pace & Trots, 2023 only. I floor the BSP.

AUS Fave Win%.jpg

At 2.0 odds, 56 started and 28 won so the expectation matched the actual. Doing a cumulative on the difference between these 2 values down all the individual BSP`s looks to point to an under performance till about 3.4 when the trend reverses

[elofan0]

http://www.adrianmassey.com/web1/pages/racfav.php

[ShaunWhite]

Your assumption is correct Card37. But I'll try to explain it from a different angle because the formulas make it seem more complicated than it is. You make the assumption for each bet that it won as often as the odds imply.

eg. Bets at 4.0 should win 100/4 % of the time, so you assume they all won, but you only take 100/4 % of the payout. 1/4 of them winning, or all winning and taking 1/4 of each win is the same.

List all your bets, sum the amounts each 'should' have won (* 100/price), sum your actual pl, the %age difference is your EV.

Two other links in addition to the one from elofan0

https://www.sbo.net/horse-racing/measur ... se-racing/
https://www.omnicalculator.com/statisti ... cted-value

[Cards37]

Thanks everyone. I should clarify I'm applying this to tennis, so only two runners, and I'm laying favourites, so they are all below $1.90. I was tinkering with the strategy in one of Peter's videos of laying favs then trading out at a defined point. It's not doing well so I wanted to quantify how much is due to variance (quite a bit I think at the moment, favourites are outperfoming expectation significantly).

[Anbell]

Hi,

I'm trying to build a few simple calculations into my tracking system to highlight variance to help shed some light on just how much impact it is having. I want to reinforce some first principles to make sure I'm not fundamentally misunderstanding.

So if a favourite is at $1.50, it has a 66.67% chance to win. If it wins, it has actually performed 0.33 above expectation. If it loses, its 0.67 below. Depending on the actual outcome, record either of those results. If the book is 100%, in the long run across a large number of events, the sum of all these should be 0. If the total is positive, favourites are winning above expectation. If its negative, favourites are losing more than expectation. Is this correct?

I have a small sample size so far of about 50 events, the running total is 3.714, which says favourites are running well above expectation.

I might be making a fundamental mistake here so would appreciate advice, thanks!

I dont think it is quite correct to say that if something has a 66% chance of winning, and it wins, and therefore it has overperformed by 33%

you might be able to say that after 100 events, but not one

[weemac]

Every fav could win but if enough of them look in trouble at any point, the final result needn't concern you. I'm sure you know that, but it seems you're not measuring that.

[ShaunWhite]

Every fav could win but if enough of them look in trouble at any point, the final result needn't concern you. I'm sure you know that, but it seems you're not measuring that.

If you're referring to trading out then you can calculate the EV on both the entry and exit. Helps to identify where you're making money or losing it instead of using the market total which doesn't tell you much.