Compounding Betfair Balance Questions

[ShaunWhite]

Within the BSP odds range 20-30 there are some individual odds that appear to be be "less likely" to win than others in that range.

Perhaps a better starting point than randomly picking out selections but probably not as much fun...

This seems like just another way to randomly select runners. There's no reason for a specific bsp to constantly outperform any other.

If anything you'd want to select the worse performing prices and hope for mean regression. Same with the lottery, it's the worst performing numbers you want not those that have already outperformed randomness.

Not a dig Athos, more a comment on if it's best to follow a streak (Eg of horses losing at 28.4) or to oppose streaks expecting reversion. But that gets us tangled in the usual clash between the gamblers fallacy and the law of large numbers. Past results don't matter but efficiency ultimately happens. Eg 2.0 horses winning 50% of the time. 10 straight losers shouldn't affect the 11th runner, but over enough it catches up to 50/50.

If this thread turns into a debate about how to find order (even short term) in the overall chaos then it'll be no bad thing.

[Atho55]

It was selected because it`s out performing it`s implied probability so nothing random about it. This is reflected in the returns since 2015.

28.4.jpg

Further along the initial screenshot is 29.4 which has a win rate % considerably greater than 28.4.

This is also reflected in the returns from 2015.

29.4.jpg

Not sure how many years you would need to find some sort of parity with the implied probability but in the meantime I know what looks to be the better bet moving forward from this point in time which at the end of the day was a pointer for the OP.

[ShaunWhite]

It was selected because it`s out performing it`s implied probability so nothing random about it.

Isn't the fact a particular price is outperforming probably in that sample random?

For me the test would be that the key values hold up on an in-sample test. Eg Find a parameter that works in 2018 and seperately see if it's also true in 2019. And then perhaps for 365 randomly chosen days. The danger of using your entire sample for testing is that some combination of values/parameters always fits.

That aside, how do you feel about the general issue of following or opposing diversion from the mean? From your approach I take it that you feel anomalies will continue, other opinions I've heard (and I think PeterW is one of them) would be to have a contrarian approach and assume that an anomaly without an underlying reason will eventually revert.

[Atho55]

Why are you trying to look for something to disprove. All the post I made was looking to do was to point the OP in a direction that on past results looked most likely to succeed going forward. In the odds range 20-30 it had lost the fewest No of times.

Why not look at your own data and say *yes that`s a reasonable choice" or "try this instead, I think it`s better"

It`s lost fewer times, no idea what the reason is and to be honest, not particularly interested. It just has. Be happy to review your findings.

[ShaunWhite]

Why not look at your own data and say *yes that`s a reasonable choice"
Be happy to review your findings.

Here you go then, I've only got GB 2013-2017 in Excel format and can't be bothered to query the SQL for racing after that. But it overlaps with your sample. In this 5 year period there were 280 horses with a BSP between 28.40 and 28.49 and 11 won.

Avg 1 qualifying horse a week. I don't need to comment on whether I think that's a reasonable choice as a fun betting proposition.

You asked for findings, I included commission at 2% but that only came to £53 laying and £60 backing. How did you get to +8 grand?
You've made me question my data now, I haven't got any winners in that price range between Sept2015 and July2017, does yours tally with that? Mine obv doesn't include IE it's just GB as I said. Try not to be defensive, I'm just trying to understand this because I'm not seeing it.

Screenshot_10.png

[Atho55]

If the OP is betting now, why would I be looking at UK & IRE data. I looked at Aus...

[ShaunWhite]

If the OP is betting now, why would I be looking at UK & IRE data. I looked at Aus...

Last 12 months of Aus horseracing had 44 in that price range and none of them won.

Less than 1 a week so not much fun, but in that time period at least, profitable.

[Atho55]

These are the figures I get. A few more than 44 potentially making it slightly more fun.

28.4 Count.jpg

Reducing the timeframe starts to make other odds within the 20-30 range nearly as attractive.

Are we looking at long term "value bets" here should the win / lose ratio remain roughly the same going forward?

[HampshireIan]

'Honey, i'm home'!!!

Time for another go at this then for shits & giggles.

I'll try and put up the bets pre race

Newcastle 14.45 - Laying Electric Missus - Odds 15.5 @ £3.30

[ShaunWhite]

Newcastle 14.45 - Laying Electric Missus - Odds 15.5 @ £3.30

I give you a 93.55% chance of getting that one.

[HampshireIan]

I give you a 93.55% chance of getting that one.

I'm rich, i'm rich

Ok, moving on:
Balance = £53.13
Newcastle 15:20:
Laying Lincoln Park - Odds 16.0 @ £0.50 and Odds 16.5 @ £2.87

[HampshireIan]

Balance = £56.38
Newcastle 15:55:
Laying Iva Reflection - Odds - 19.5 @ £2.89

[HampshireIan]

Balance = £59.13
Newcastle 16:30:
Laying Gorgeous General - Odds - 20 @ £2.95

[HampshireIan]

Balance £61.83
Newcastle 17:05pm
Lay Makawee - odds 20 @ 3.25

[HampshireIan]

Balance £65.02
Newcastle 17.40pm
Laying Flames Of York - odds 27 @ £2.49