Drifter/Steamer statistics

[renoo]

I know there was an official Betfair report from 2008 or so which implied you would have been far better off backing horses where BSP was significantly higher than average matched price on a runner.
However my personal impression without having much data to back it up is opposite, does anyone know if this is still generally the case and what would be the easiest way to analyse statistics like that? Like for example how often did BSP 10 win, if odds drifted more than 20% etc.
Thank you

[ruthlessimon]

I queried this with another user a while back, because it makes sense that it should be an edge (if auction market theory is valid).

If I understand it correctly, this is what BF suggest is a value back (screenie taken at exact offtime):



My source said it wasn't an edge (but maybe he would ), & also I'm not quite sure how he set up the test.

Unfortunately, I don't collect inplay myself - maybe I should

[ruthlessimon]

If the technical definition of "fair value" is 2.46 - 2.5; then considering (relatively) little to no volume traded @ 3.0 makes it a completely "unfair" lay price.

Although markets are extremely complex beasts; people forget that!

[LinusP]

Easy enough with the Betfair sp data:

http://www.betfairpromo.com/betfairsp/prices/

[northbound]

I went through my pre-race snapshots (Oct2017-Apr2018) and simulated bets on all horses with the following criteria:

- Bet at scheduled race start time (if the race started 5min later, my data doesn't know)
- BestBack price 10-15% bigger than WeightedAverage price
- BestBack price between 2.20 - 4.00
- Bet £1 on the horse to win the race
- Subtract 5% commission on winnings


Here are the total profit/loss figures:

UK Races P/L (Total): -£0.08
IRE Races P/L (Total): £5.54


Here are the total profit/loss figures for the better performing race types:

UK Races P/L (Flat Handicaps): £4.25
UK Races P/L (Jumps Handicaps): £4.63
IRE Races P/L (NHF): £4.77

[northbound]

Here are the results for what the original poster suggested:

- Bet at scheduled race start time (if the race started 5min later, my data doesn't know)
- BestBack price 20%+ bigger than WeightedAverage price
- BestBack price between 6.00 - 20.00
- Bet £1 on the horse to win the race
- Subtract 5% commission on winnings

The profit/loss figures for these higher priced selections are really lousy:

UK Races P/L (Total): -£140.74
IRE Races P/L (Total): -£115.50

[Euler]

Backing a drifter doesn't make much sense to me as a price can drift for many reasons. But laying a steamer does.

[ruthlessimon]

The profit/loss figures for these higher priced selections are really lousy

Was the sample size of trades for both strategies meaningful?

[northbound]

Was the sample size of trades for both strategies meaningful?

All UK / IRE races from Oct 2017 until Apr 2018.

I don't have snapshots prior to that.

[ruthlessimon]

Was the sample size of trades for both strategies meaningful?

All UK / IRE races from Oct 2017 until Apr 2018.

I don't have snapshots prior to that.

It's just that on my data a 20% drift is extremely rare; let alone from the WAP, which would have formed very late!

[ruthlessimon]

Backing a drifter doesn't make much sense to me as a price can drift for many reasons. But laying a steamer does.

It makes sense doesn't it, except on my pre-off data - laying a drifting fav sucks massively (exit @ 0mins).

April:


May (so far):


Interesting that the start of both months saw a blind lay strategy doing well(ish). & then both pretty much down the same amount as the previous

[Euler]

WAP is always going to be skewed heavily to the last couple of minutes as that is when most of the money arrives. So I'd imagine a variance related to that is a little skewed also?

[northbound]

Was the sample size of trades for both strategies meaningful?

All UK / IRE races from Oct 2017 until Apr 2018.

I don't have snapshots prior to that.

It's just that on my data a 20% drift is extremely rare; let alone from the WAP, which would have formed very late!

20% drift on a 6.00 - 20.00 horse doesn't look so uncommon in my data.

I'm talking about snapshots taken at scheduled start, so perhaps the data at the actual off is a bit tighter as more money arrived in the next few minutes of a late race, thus reducing the gap between WAP and BB.

Here are a few random horses from my sample:

2018-03-20 Newcastle Velvet Vision WAP 5.39 / BB 7.60 41%
2017-12-27 Wolverhampton Jumping Around WAP 15.22 / BB 18.50 21%
2017-10-20 Redcar Essential WAP 13.11 / BB 16.5 25%

[ruthlessimon]

2018-03-20 Newcastle Velvet Vision WAP 5.39 / BB 7.60 41%

I'd calculate that as a 5% drift, not 40%!

[northbound]

2018-03-20 Newcastle Velvet Vision WAP 5.39 / BB 7.60 41%

I'd calculate that as a 5% drift, not 40%!

BB 7.60 / WAP 5.39 = 1.41

BB is 41% bigger than WAP in my mind