Lost the edge or just probability?

what do you mean by 'a sharp downturn'?

"sharp downturn"=done his nuts ?

0n the last 118 races I've only had 27 winners. The win loss ratio has remained as before i.e. 1:0.9

TTT

I have done my nuts but even only doing 1% of the bank on each race I've taken a bit of a bath.

I have no edge at the moment, it's the weather that is at fault. Same happened last year in feb. I'd suggest reducing you stakes and waiting for a month. You might find it recovers.

I'd suggest cutting back on stakes until you see an upward trend again.
Sometimes,even a 1,300 race sample isn't enough to prove conclusively that you had an edge in the first place.
Good luck!

Thanks TTt. Do you have an idea for the number of races you need to identify the edge exists? I presume it has something to do with the size of the perceived edge.

AW going does change ....and SLOW going will effect results......

Each course on AW is different.......so your "edge" maybe only works on STANDARD????

PSII is spot on IMO......

I like the going varying as if you do have an edge it makes choice of bets far easier.........

SOUTHWELL has been main course lately ....and all been on SLOW .........this probably is why you are not doing well.

GL with bets

PSII - That says that the markets changes with the weather - which sounds interesting. Can I ask why have you come to that conculusion?

I think DFCIronman has answered it - many thanks

I guess it does.If your edge was 90%,you would expect to get in front pretty quickly,so only a small sample would be required.
It could be a statistical blip,or a reversion to the mean.

Just from experience really. Two years ago the country flooded during the summer. What racing that did go ahead I made nothing on, and I made nothing for the first month after it returned. Same happened last year, I had a losing february after the snow. And the same is happening again now. I think the problem is two fold: after a long lay off form is very uncertain and insiders, or just people with good models, pick you off. Secondly, liquidity drops and you end up pro vs pro. No one's really going to win in that situation.

That's my take on it anyway.

Oh, after we had the floods I found I was losing masses of cash when the going was heavy (which was common) and doing five otherwise. So I agree with DFC that going is an important factor.

Thanks PPII - very plausible

But why would the going be a factor ? If a flaw can be identified, then you have a new edge surely.

It Happens

I had a system which went from a £60 Bank to +£8,000 in 1000 races/3 Months :)

( See chart ) http://tinypic.com/view.php?pic=2w658wj&s=6

After 1250 races it was upto +£14,500

at +£12,000 Champagne was cracked open and congratulations were modestly accepted.....

Then it just started going tits up and four months later I glumly admitted to myself that the
system had lost its edge and switched it off :( About £6k in profit....... Back to the drawing board

I've lost the chart for the whole 2300 races but here's the first 1580
http://tinypic.com/view.php?pic=24bu26h&s=6

The down trend just continued with lower stakes



I've had various edges in Spread betting financials, poker and Betting Exchanges, and in my own experience, they mostly ( but not always ) diminish or disappear...... you have to get it while you can.



If you think you have an edge then use a very aggressive staking plan with a small starting bank is my advice, because you never know how long it'll last.

Escapee

That curve looks remarkably like mine - the difference being that my down phase is worse!

Post

Ooops pressed wrong button


Artisan, Post a chart of it on tinypic.com and paste the linnk here ( please )

if the strategy has good logic - stick with it

If it's horse then forget it at the moment, I have stopped my bot from trading them because the last three weeks the results are not statistically normal and they have just been costing me money, luckily a couple of other markets have been over performing so it's not been too bad a problem.

Low grade racing has a higher "take out" from insiders. It thus makes making a profit even harder than usual.

We are obviously experiencing much more low grade racing because of the weather.

Someone good at statistics will be able to say how often a 27% strike rate from 100 runners is going to happen for an event with 53% probability (if that is what your true probablility is from a larger sample, 1300 is not very large)

Kevin Pullein was very good at doing that, when everyone would be looking for a new reason to explain a short-term trend in football he would show that those patterns happen once or twice every season. But i'd also concur with those recommending to sit back due to the current conditions

You'll never know for sure. All you can do is work out mathematically what the chance is that you even have an edge based on your past results. You'll see that if you work out this chance using data from before things started to go wrong, it will be much higher than if you use all the data.

This is just like a sports event where odds can swing wildly after something important happens.

The above assumes you're just going by your results (not any underlying theories of how and why you make or lose money).

If you do a simulation of 10,000 coin tosses with a biased coin, where you constantly bet a fixed stake on heads which occurs 50.5% of the time, and put the results in a graph for each simulation, you'll see the huge differences in the graphs.

If the biased coin landed on heads 55% of the time the charts would look much more alike.
If I showed you just one chart telling you this is an experiment with a biased coin, but not telling you what the edge was, it would be practically impossible for you to work out with a high degree of accuracy if the edge was small, and easier if the edge was large.

However, if all the information you have is a chart and P/L figure, it can still be misleading.
For instance, if you saw a series of 100 consecutive winning bets, it may look amazing, but it could just be someone laying at 1000.

If you want to check out the approach to this there are articles covering it at smartersig.com

In regards to the OP.

One common Problem is that many Selection criteria one might come up with that produce good overall results arent profitable every year !

Considering the frosty weather it comes probably down to that and the recent going issues.

Besides that betting on the AW Surface is always harder than on turf.

Over time i got the impression that there is less money to be made on the Artifical Surface than on turf..... Of course there is also some good stuff around that piles up the monet quite nicely :)

Cheers

Sample of 1000 is accurate to +/- 3% at a 95% confidence level

If you want to halve that to 1.5% you'd need a sample of at least 10,000

Not at statistician so happy for someone to correct me on this but I remember this from my old day job

joshuag 17 Jan 16:28
If you want to check out the approach to this there are articles covering it at smartersig.com

Is it possible to read these articles, or samples somewhere? Most of it isn't of interest to me as I'm not involved with horse racing.

@ escape

Well one more thing,..... if you made 8000 Bucks over 1000 Races in 3 (!) months than there is no wonder that things take a downswing now cause thats not normal !!

I mean the Racing markets on here are pretty bad but i guess they arent that bad !!

I consider it normal that the Market takes back a big chunk of your excellent Profits.

For example:

Last Turf Season i had one System running that gave me a 120% Roi over the Summer months,... since then its dipping away tiny bits for the last 5 months,.... the criteria dont seem to work really well on the sand and it was to much money to fast anyway to consider it "expectable".

1000 races is nowhere near enough to make black and white conlcusions...

if you cant afford the swings that you will see on a new system during the first 5000-10000 races bet smaller, and run your system longer.

do the statistical analysis on your data also - but generally i dont think most people on here run thier systems into big enough numbers at affordable stakes.


Jan is always tough - epsecially last few years.

many reasons some alluded to already

but I think especially there are more and more smart people fishing out the casual punter. so now the smart people need to beat other smart people to make it pay.

I have little doubt 5-10 people provide 30-40% of the real liquidity on this exchnage - and winning in that market is going to get tougher and tougher

I speak of horse racing

Difficult to be certain of anything but for what it's worth i'm an advocate of needing to know why something works irrespective of any figures over any worthwhile period , whilst i'm a believer in stats to support any system that's all it should ever do , never drive the system in other words.

So two key points for me would be , ask myself to write a paragraph explaining why it works and then remind myself that statistics don't give answers they only raise questions.

good luck and keep us all posted.

Hayden,

Are you saying that if you discovered something was profitable in the past, monitored results for a while and found it was still profitable (consistently), you wouldn't act on it if you didn't understand why?

In this situation some people would certainly 'make up' a reason that seems plausible, as is normal practice. it is very difficult to simply accept you don't know something.

yes that's exactly what i'm saying tbh , I would be comfortable explaining why my system works by logic and sensible reasoning alone and then offer the stats as a support if required , think you'll find that most systems developed with stats alone are suspect in that they have been " tampered " to fit the requirements at some stage and I don't mean that in a cynical sense it's just that the developer isn't even aware they've done it in most cases.

Err how can you know something works irrespective of figures?

Just remembered a useful check whether a system is still valid and just going through a statistical abberation is to assess whether the number of 2nds is in line with what you expect or whether they too have fallen away

the only system stats you need are your Pand L

If you're in the bath pull the plug out

there was a full moon on 11th Jan ;)

you can model the probablility of a losing streak in Excel using monte carlo simulation.

Since you have a known sample size and a win loss ratio over time. you can then run the sim to establish what the percentile chance of a losing run of x bets is.

More simply you an look back and see if you have had a heater somewhere with an equal but opposite trend, (or very similar), which would suggest it to be a feasible poor streak.

Otherwise you would have to know a lot more about your trading methods to find places tha tmay have changed.

Thanks for the replies - all very helpful.

So it seems to me that one or more of a number of things could be happening here as follows:

1. I didn't have an edge in the first place, my back testing sample wasn't large enough, and/or the underpinning logic of the system wasn't sound.

2. The edge is still there, but I've just experienced a probabalistic losing streak.

3. Conditions have changed (weather, AW tracks etc).

4. The market has changed for some other reason (new entrants with new systems, or old entrants withdrawing their systems).

Following TJM's excellent suggestion, I've done a simple excel Monte Carlo simulation, which to me pretty conclusively shows that this can't be a probabalistic losing streak. With a simulation set of 10,000 races the largest draw down I found over 500 simulations was 38 points, when I've just experienced 79 points (using 53.3% strike rate, 1:0.9 win:loss ratio). Although, some might say more simulations are required this to me shows that the probability of it being a probabalistic losing streak is very very small.

The simulation also showed me (I think) that for my strike rate and win:loss ratio between 1,000 and 2,000 races should be enough to verify the edge.

So I'm left with, did I have an edge in the first place, or what's changed?

I'm as sure as I can be that I did have an edge - the unlying logic was sound, and I repeatedly tested and analysed it with a testing sample set of over 10,000 races. The results, up until the downturn, pretty much matched the anticipated results from the testing.

I looked at the performance both in test and in live for AW. I've known from previous analysis that AW, although profitable, does not perform as well, but certainy not as badly as I've just experienced. What was interesting was I looked at the last two January periods and also the floods - and historically these have definitely been the systems weakest.

So could it be that the weather really is to blame, and maybe when the weather is bad and there is little in the way of racing, the exchange changes its characteristics?

As recommended I'm reducing the stakes and hanging in there - the last three days appear a lot more "normal".

Thanks for the replies - all very helpful.

So it seems to me that one or more of a number of things could be happening here as follows:

1. I didn't have an edge in the first place, my back testing sample wasn't large enough, and/or the underpinning logic of the system wasn't sound.

2. The edge is still there, but I've just experienced a probabalistic losing streak.

3. Conditions have changed (weather, AW tracks etc).

4. The market has changed for some other reason (new entrants with new systems, or old entrants withdrawing their systems).

Following TJM's excellent suggestion, I've done a simple excel Monte Carlo simulation, which to me pretty conclusively shows that this can't be a probabalistic losing streak. With a simulation set of 10,000 races the largest draw down I found over 500 simulations was 38 points, when I've just experienced 79 points (using 53.3% strike rate, 1:0.9 win:loss ratio). Although, some might say more simulations are required this to me shows that the probability of it being a probabalistic losing streak is very very small.

The simulation also showed me (I think) that for my strike rate and win:loss ratio between 1,000 and 2,000 races should be enough to verify the edge.

So I'm left with, did I have an edge in the first place, or what's changed?

I'm as sure as I can be that I did have an edge - the unlying logic was sound, and I repeatedly tested and analysed it with a testing sample set of over 10,000 races. The results, up until the downturn, pretty much matched the anticipated results from the testing.

I looked at the performance both in test and in live for AW. I've known from previous analysis that AW, although profitable, does not perform as well, but certainy not as badly as I've just experienced. What was interesting was I looked at the last two January periods and also the floods - and historically these have definitely been the systems weakest.

So could it be that the weather really is to blame, and maybe when the weather is bad and there is little in the way of racing, the exchange changes its characteristics?

As recommended I'm reducing the stakes and hanging in there - the last three days appear a lot more "normal".

Well done, sound analysis, and well worth bearing in mind the implications of bad weather and cancelled meetings for the future

I have found my edge significantly diminished since september last year. It first started creaking in june following a long period of stability. Can't really explain it. It could be seasonal factors but these have not had an impact before. More likely just the fact that the markets change over time - people with new strategies etc.