When do you know that a strategy works ?

[THENUTS]

Good Evening,

I'm trying to understand when a strategy can move away from testing into a proven positive expected value situation that can be scaled and not just something that is waiting to revert to the mean with no edge

I have a fully automated (excel) horse racing in running lay strategy that has now been applied to 504 races since 19th March for flat stakes of £9

504 races has produced a return after commission of £431
The edge (cum profit/cum liability) has leveled out at 2.1% over the last hundred or so races

I have 2 question's for anyone that has experience of these kinds of fully automated situations ( Peter Le , Dallas , Euler and others ) :

1) 504 races over almost 1 month seems encouraging but do i need to wait for 1000 or 2000 races before increasing stakes ?

2) Once i establish that it is a winning strategy , how do i scale it ? I'm thinking add £1 for every £500 it makes but i really have no idea

I have attached a graph showing progress so far.

Any feedback which would help me move this forward would be much appreciated

layedgegraph.xlsx

[Bluesky]

I'm sure the experts here will give you some great information. The only think I wanted to ask you was have you watched much of the trading taking place on the ladders during the actual race? You will often see that there is very little money available at various prices, and it is withdrawn very quickly.

I doubt you will have a problem at your current stake level and I am sure you will be fine to increase for a while. I do think however that stake size will become a problem for most of the typical races and will be something you will eventually have to take into account, if indeed you have found a winning strategy (your system certainly looks like it might be a good one).

If no one does give you any good suggestions (I'm sure they will), then take a look on line for some of the equations that traders can use to provide confidence levels for their systems. You sound like a math/programmer type person to me so I am sure you would have no difficulty understanding the formulas available to give you a good idea how likely your system is to be a success. Obviously the more trades you have, then the more confidence you can have in the result produced by the above mentioned equations.

Good luck to you and I hope things work out, sounds to me like your on the right track and you seem to be approaching things in the correct manner.

[THENUTS]

The only think I wanted to ask you was have you watched much of the trading taking place on the ladders during the actual race? You will often see that there is very little money available at various prices, and it is withdrawn very quickly.

Yes , have watched most ladders as the bet gets triggered. Plenty of leway to increase stakes at least 10 fold before matching would become a problem

[Anna List]

Good Evening,

I'm trying to understand when a strategy can move away from testing into a proven positive expected value situation that can be scaled and not just something that is waiting to revert to the mean with no edge

I have a fully automated (excel) horse racing in running lay strategy that has now been applied to 504 races since 19th March for flat stakes of £9

504 races has produced a return after commission of £431
The edge (cum profit/cum liability) has leveled out at 2.1% over the last hundred or so races

I have 2 question's for anyone that has experience of these kinds of fully automated situations ( Peter Le , Dallas , Euler and others ) :

1) 504 races over almost 1 month seems encouraging but do i need to wait for 1000 or 2000 races before increasing stakes ?

2) Once i establish that it is a winning strategy , how do i scale it ? I'm thinking add £1 for every £500 it makes but i really have no idea

I have attached a graph showing progress so far.

Any feedback which would help me move this forward would be much appreciated


layedgegraph.xlsx

You haven't told me a lot about your system, and I'm fine with that, but it does mean that I'm going to have to base some of my comments on guesswork. I hope that this is OK with you.

I'm guessing that your in-running lay system has a strike rate of somewhere in the region of 80%. Let's suppose, for the purposes of this post, that it is exactly 80%.

Based on 504 races and an 80% strike rate, your margin of error is approximately 3.5% at the 95% confidence level. Therefore, you can be 95% confident that the actual strike rate of your system is somewhere between 76.5% and 83.5%.

If it's actually 83.5%, that's fine because the system will be even more profitable in the future. If it is only 76.5%, what then? Does this mean that your system won't be profitable long term? If this is the case, you need to perform some more testing because the margin of error decreases with testing.

Based on your strike rate, your longest losing run (LLR) ought to be about 4.

The length of the LLR that you are likely to encounter is dependent upon your system's strike rate and the number of bets that you have placed. The higher the strike rate, the lower the LLR. The higher the number of bets that you have placed, the higher the LLR.

If you allocate 1/4 (25%) of your betting bank to each bet, it means that the next time you encounter a losing run of 4, your bank will be completely lost. Therefore, in order to continue betting you will need to allocate less than 25% of your bank to each bet.

That's not the end of this though.

As you bet more, the longest losing run that you could encounter increases.

That's still not the end.

It isn't just the LLR that can cause the loss of your bank. Lots of smaller runs of losses which are close together can do as much, if not more, damage to your bank.

For the above reasons, in order to determine what liability of the bank to allocate to each bet, multiply the LLR by a factor and then divide the result into the bank.

If I have a new system and I don't trust it as much as I should, I multiply the LLR by 5. As I come to trust the system more and more, I start to reduce the factor slowly but, I never reduce the factor to less than 3.5.

The issue with testing is that, sometimes, a system doesn't convert from testing to real life. For example, in testing, you might assume that all of your stake will be matched at the required odds whereas, when you go live, this may not be the case.

Also, scaling up may not be as easy as you think. In-running liquidity, in general, is in short supply. At the specific odds required, it's possible that there is no liquidity whatsoever.

Also, pre-race, waiting in a queue to get matched is far less of an issue than it is in-running. You can't just dump some cash into the in-running market and wait for it to get matched. Well, you can, but I wouldn't recommend it.

Maybe you have though about the above issues, and more besides.

Hope this helps and, if you require further explanation, please don't hesitate to contact me.

Anna

[PeterLe]

Take a look at this thread
http://www.betangel.com/forum/viewtopic.php?f=2&t=12099
Regards
Peter

[THENUTS]

I'm guessing that your in-running lay system has a strike rate of somewhere in the region of 80%. Let's suppose, for the purposes of this post, that it is exactly 80%.

Based on 504 races and an 80% strike rate, your margin of error is approximately 3.5% at the 95% confidence level. Therefore, you can be 95% confident that the actual strike rate of your system is somewhere between 76.5% and 83.5%.

If it's actually 83.5%, that's fine because the system will be even more profitable in the future. If it is only 76.5%, what then? Does this mean that your system won't be profitable long term? If this is the case, you need to perform some more testing because the margin of error decreases with testing.

The issue with testing is that, sometimes, a system doesn't convert from testing to real life. For example, in testing, you might assume that all of your stake will be matched at the required odds whereas, when you go live, this may not be the case.

Thank you for your comprehensive response Anna , interesting methodology which i shall certainly use going forward.

I like how you managed to get the strike rate so close in your assumption. I calculated at the outset that i needed around 80% to break even and am acheiving 83.2% currently

Also, perhaps i should have been clearer when i said "testing". I'm not using practice mode , these results are actual results in the live market.
I meant testing at these relatively low stakes before scaling up to hopefully make it something really worthwhile ( i'll slowly scale to ensure i get matched as the stakes rise and monitor closely my edge % if i only start o get partially matched )

Thanks again for taking the time to respond in such detail

[THENUTS]

Take a look at this thread
http://www.betangel.com/forum/viewtopic.php?f=2&t=12099
Regards
Peter

Thanks Peter - very helpful

[LinusP]

Have a look through all of PeterLe's threads, lots of gold in there

Decided to have a look at an average horse racing strategy that I know has been profitable for 4 years, the graph below shows the profit over the past 8 months (14k races).

screenshot.png

You can see that for almost 2 months it was losing money week on week and another 2 months were it was breaking even. There is a lot of a luck involved in some strategies as I doubt I would have continued this had I started it Nov 2016. Horse racing can be so volatile that edges/value can sometimes disappear overnight. I wouldn't be surprised if the downtrend in the strategy above was due to someone else doing the same thing as its very simple.

When it comes to stakes I actually work mostly with £2 and its rare for me to use anything above but I would recommend looking into kelly criteria and sharpe ratio.

[Anna List]

Have a look through all of PeterLe's threads, lots of gold in there

Decided to have a look at an average horse racing strategy that I know has been profitable for 4 years, the graph below shows the profit over the past 8 months (14k races).


screenshot.png

You can see that for almost 2 months it was losing money week on week and another 2 months were it was breaking even. There is a lot of a luck involved in some strategies as I doubt I would have continued this had I started it Nov 2016. Horse racing can be so volatile that edges/value can sometimes disappear overnight. I wouldn't be surprised if the downtrend in the strategy above was due to someone else doing the same thing as its very simple.

When it comes to stakes I actually work mostly with £2 and its rare for me to use anything above but I would recommend looking into kelly criteria and sharpe ratio.

Linus

For all sorts of reasons, thanks for posting the above.

There's much food for thought in that there post.

Anna

[THENUTS]

There is a lot of a luck involved in some strategies. Horse racing can be so volatile that edges/value can sometimes disappear overnight.

SInce i started this thread my +£431 has become +£280 in 2 days..........bit of a blow but all the more interesting to see your graph over a longer period and the long periods with losses/no gains

I shall persevere

[Anna List]

There is a lot of a luck involved in some strategies. Horse racing can be so volatile that edges/value can sometimes disappear overnight.

SInce i started this thread my +£431 has become +£280 in 2 days..........bit of a blow but all the more interesting to see your graph over a longer period and the long periods with losses/no gains

I shall persevere

Hi TheNuts

Firstly, let me say how sorry I was to hear about the downturn in the fortunes of your system. I take no pleasure in hearing about the ills of others, believe me. I've had enough of my own.

I tend to be a little critical when it comes to all things gambling. I find that it's safest to err on the side of caution in such matters so please forgive the following comments.

In your original post, you wrote that your edge was 2.1%. I calculated that your margin of error was 3.5% which, when subtracted from your strike rate, would take you below the threshold required to maintain a profit. I believe that you calculated that this threshold was 80%?

When a system has a downturn, all sorts of questions are asked of it such as: Why is the system going through a loss-making phase? Why now? When will it end? Will it end? Yet, when the system is going through a winning phase, in general, not one word is uttered.

It is entirely POSSIBLE that you have a break-even or a losing system which just went, for whatever reason, through a profitable phase. Unfortunately, that phase is now over. It is also entirely POSSIBLE that you have a winning system which is now going through a losing phase. Which of these statements is true is unclear at this time.

Basically, imho, your edge of 2.1% was too low and your margin of error was too high to contemplate going live. In your position, I'd put the system back in test mode and monitor. That said, sods law states that if you do, you will miss out on the best betting day that you will ever have. Such is life.

Hope things work out for the best for you.

All the best.

Anna