I thought I had finally found a winning strategy for UK horse races. From April to July I was winning every week and only had 19 losing days out of 116. I was only winning on 55% of races, but my wins were always bigger than my losses.
Since 1st August though, I've only had 5 winning days. Losing days are losing more than winning days win. And I'm only winning on 46% of races.
Does that mean that my strategy was just luck?
Or am I just going through a losing period. In 18 days I've lost 12% of what I won in April to July.
Or does UK horse racing change in different months and I should put my strategy aside and try it again next year?
After years of having more downs than ups, I thought I had finally cracked it. But now I'm wondering.
My strategy was 100% automated because I have never been able to trust myself trading manually.
All thoughts welcome please.
What's gone wrong?
-
ShaunWhite
- Posts: 10851
- Joined: Sat Sep 03, 2016 3:42 am
No need to panic, it happens. Its just 2 weeks out of 4 months. But what you need to do now is to reduce your stakes, maybe even to min stakes. Not because you hope to beat any variance by guessing what next week will hold, but to hang onto your profits while you assess the next few months. If your strategy has actually faded then you'll retain most of your previous gains, if it shows an upturn then you can scale up again with some renewed confidence. Making steady money is about not giving it back as much as it's about finding a rich vein.
Reducing your stake will help relieve the mental pressure too, it's mentioned a lot by discretionary traders, but they have the luxury of each day being a fresh start. Often with automation you can take weeks developing a stragegy and then it's weeks until the averages show their true colours, it's psychological pain in bullet time, death by 1000 cuts, and as damaging to your discipline and a manual trader going on tilt.
So, drop stakes, monitor, reassess after a month or so. In the meantime work on something else totally different.... And generally, allow yourself an R&D budget, take 10% or more of the gains you've made so far, and use that as no pressure money you can burn on testing/trialing things or just gaining intel. It's much harder when everything you do matters so allowing yourself an R&D budget eases that.
Reducing your stake will help relieve the mental pressure too, it's mentioned a lot by discretionary traders, but they have the luxury of each day being a fresh start. Often with automation you can take weeks developing a stragegy and then it's weeks until the averages show their true colours, it's psychological pain in bullet time, death by 1000 cuts, and as damaging to your discipline and a manual trader going on tilt.
So, drop stakes, monitor, reassess after a month or so. In the meantime work on something else totally different.... And generally, allow yourself an R&D budget, take 10% or more of the gains you've made so far, and use that as no pressure money you can burn on testing/trialing things or just gaining intel. It's much harder when everything you do matters so allowing yourself an R&D budget eases that.
Here is a very simplistic Beta-Binomial analysis to illustrate the evidence that something may have changed.
Assume each trading day is a winning day with probability (p), and begin with a uniform prior (no preference either way):
p ~ Beta(1,1)
**April-July**
- From April–July:
- Posterior mean: 98/118 ~ 83.1%
- Approximate 95% credible interval: 76%–90%
**August**
- For August:
- Posterior mean: 6/20 = 30%
- Approximate 95% credible interval: 12%–52%
The 95% credible intervals do not overlap which suggests something has changed!
Bottom line
This is not likely to be just a routine fluctuation. The sensible response is to pause or sharply reduce stakes and investigate, rather than assume it will recover (As @ShaunWhite advised).
**Important Caveats**
- Winning days may not be independent; several bets can share the same market conditions.
- Also, profitable/unprofitable days discard information about the number of bets and the sizes of wins and losses.
- This analysis is a useful warning signal, not a complete evaluation of the strategy.
Assume each trading day is a winning day with probability (p), and begin with a uniform prior (no preference either way):
p ~ Beta(1,1)
**April-July**
- From April–July:
- 116 days
- 97 winning days
- 19 losing days
- p(apr-jul | data) ~ Beta(97+1, 19+1)
- Posterior mean: 98/118 ~ 83.1%
- Approximate 95% credible interval: 76%–90%
**August**
- For August:
- 18 days
- 5 winning days
- 13 losing days
- p(aug | data) ~ Beta(5+1, 13+1)
- Posterior mean: 6/20 = 30%
- Approximate 95% credible interval: 12%–52%
The 95% credible intervals do not overlap which suggests something has changed!
Bottom line
This is not likely to be just a routine fluctuation. The sensible response is to pause or sharply reduce stakes and investigate, rather than assume it will recover (As @ShaunWhite advised).
**Important Caveats**
- Winning days may not be independent; several bets can share the same market conditions.
- Also, profitable/unprofitable days discard information about the number of bets and the sizes of wins and losses.
- This analysis is a useful warning signal, not a complete evaluation of the strategy.
All sports are seasonal to some extent, so you'll find there are patterns within an overall season.
Not just going from flat to jumps, but, for example, at the early part of the season, form isn't very well understood. There's a lot of variability, but as the season trundles on, form becomes much clearer.
Not just going from flat to jumps, but, for example, at the early part of the season, form isn't very well understood. There's a lot of variability, but as the season trundles on, form becomes much clearer.