Sufficient test period

[Hammers]

I have been trialling a selection method on spreadsheet only. The action is high and the results very encouraging. My question to the forum is one that may well have been asked by someone else previously so I apologise if this is repetitious -

How many bets or what period of time is sufficient to satisfy ones self that the test is reliable?

My figures are as follows ....

976 bets
305 winners (31.3%)
562 places (57.6%)
Win POT 28.8%
Place POT 7.1%

37 betting days - 30 winning days, 7 losing.
Average number of bets per day 26.

There is one $33 winner in the mix, 14 at between $10 and $19 and 290 winners at single figures. I say this so that the figures are not seen to be overly flattering owing to a large number of long priced winners that may not recur.

Is it time to bet with a small bank to test the water, dive in head first, or stay on shore and watch for a while longer?

I would appreciate some advice - I pick the winners okay but my execution in the past has been woeful to say the least.

Cheers,

Hammers.

[Percentum]

Hello Hammers
A few months ago I was putting tips on here using some of my earlier research. The earlier research had shown a profit of about 18% over in excess of 1 500 bets. I felt confident posting these in light of,what appeared to me, a good sample. Nevertheless it was down by about 150 units by the time I gave away boldly posting its predictions. Not sure there is a definitive sample amount. I think it is the statistical outliers that sustain most profitable systems. Though it would be nice, I'm not convinced there are too many methods out there that allow you to cruise along harvesting steady profits without ever endangering the bank.With me it's a bit like fishing - put a few lines out, drift along staying afloat and hoping the big one comes along. Anyway, congratulations if you've found a method that continues to work and congratulations on your hard work. Keeps the likes of us off the streets eh?

[crash]

Past results are a reliable indication of how the system went in the past. If a system makes sense I can see no reason not to start a trial run betting on it.

[stebbo]

Hammers,

there is one way, and only one way, to adquately test a system. Put real $ on it. Even if it's only $10 per bet, you have to put real money on the selections before you can have any confidence in them whatsoever.

Work out the maximum drawdown and double it. Work out the longest run of outs and multiply it by 4. Take the largest of these two numbers and this is how many units you should allocate. Multiply this by how much you will put on each bet and put that amount of money into a bookie account.

Then bet them all. Every single one. Until you have lots of money or the account is zero. Don't pull out if the bank is down by half or down by three quarters, or even if the bank gets close to zero. Bet them till the bank is actually zero. If you pull out before the bank is zero then you haven't given the strategy an adequate test.

For example, if your Maximum Drawdown for your existing trial is 60 units, then you need to allocate a 120 unit bank. If you wish to bet $10 per unit, then set aside $1,200 that you can afford to lose. Start betting. And don't stop until that bank is completely gone.

Cheers,
Chris.

[thorns]

I think with the high turnover method it seems to be, it would be worthwhile collecting a few more months results before putting your own hard earned on the selections. Your not going to loose anything by watching for a bit longer.

I have been down the same road as Percentum, think I have a great system, only for it to start going backwards at a fast rate of knots once I start betting betting it. If you do decide to put money on it, start with a small test bank, and see how you cope with the enevatable loosing runs that occur with any system. If with just small amounts on, you get uncomfortable with the run of outs, you are never going to be able to bet it with large sums.

But, yeah, dont hurry into it, it costs nothing to keep paper trialling for a bit longer.

[Hammers]

Thank you for the generous replies chaps.

After taking in some of your thoughts I reckon a bet size of 1% of max bank level with a strike rate of 30 odd % seems conservative enough to weather the worst of bad runs yet positive enough to enable worthwhile growth. A bank such as this would have reached $14719 today after starting with $1000 39 days ago.

Given what you guys have said I am minded to test a little further before betting real money - maybe after 2500-3000 bets.

I've tinkered with a million ideas over the years but this one is the only I have come across with the vital combination of reasonable POT and high action. I believe turnover is king - much rather 1% POT on 1000 bets than 100% POT on 2 bets.

Thanks again for the input - the strength of this forum is the collective wisdom of the participants.

Cheers,

Hammers.

[Bhagwan]

I feel this staking approach may appeal to some.


Level stakes bet to a ratio of 200:1

e.g. $1 for every $200 bank.

Adjust the unit bet for the day by dividing 200 into current bank amount up & down at the beginning of a new day.

e.g. If down to $180 / 200 = $0.90 is the unit bet for the day.

The 200:1 ratio smooths out most rapid swings , up or down.

Try not to lower this ratio too much, other wise one will find themselves placings too little on the winners & more on the losers as they arrive in sequence order.
This ratio will be doing the same thing but to a lesser extreme therefore greater POT if all pans out to our favour.

Well done on your system.
I have found that systems tested for approx 4 years seem to have a greater chance of standing up long term.

Meanwhile protecting the bank.

It will survive most horror runs.

If one has a strong run , one will do well.

With the 31% SR the theoretical run of outs in a row is approx 20 so be prepared, this is normal, it's all part of the process, so try not to be too concerned when it occurs, because your money management plan should be allowing for this.

I have found that systems tested for approx 4 years seem to have a greater chance of standing up long term.

One method of testing is to break up ones sample into batches of 150
If say one batch shows a profit & the other 2 break even or slight loss it now shows potential to continue into the future.

Another way is to break up the sample between odd numbered races & even numbered races .

If there is a large discrepancy , watch out for some surprises in the future.
Or you may feel more comfortable only backing the races, odd or even numbered that have shown the greatest potential, just for the fun of it.

What one likes to look for is consistency in the strike rate.
Once that is established, the next trick is to get value, this bit, is the thing that may lead to disappointment for most systems.

It's funny when one finally finds something that works, then the value falls away.
It's the low prices that may slow the plan to treading water so try not to bet with the TAB's, use anything else e.g. Betfair or best of TOTEs bookies but not TABs.

Another thing to try & keep in mind if using multiple selection plans, is try not to double up on selections mentioned twice, otherwise, one will cannibalise their profits long term.
We don't want to assume that because the selection is mentioned twice that it has twice as much chance of winning therefore placing twice as much on it.

Well done on your system Hammers.
Now all you have to do now, is put the rules up for all to enjoy.

Cheers.

[Hammers]

Thanks for your post Bhagwan,

I have divided my selections into each capital city and each mainland state for provincials and country meets i.e. 10 categories. Only Melbourne shows a loss and quite a severe loss - maybe something to do with the carnival.

I have broken the races into the different going types and all show a profit. I have split the selections into 8 different distance categories and all show a profit. I am hoping that this shows some sort of reliability along the lines of what you were talking about Bhagwan with the odds/evens and so on.

The only anomoly was days of the week - Sunday was losing quite badly, Monday losing a little and the other days healthily in profit. But yesterday was a red letter day and Sunday is now back near square.

I think your 4 year test period will try my patience. Given that I am finding an average of 26-27 bets per day I will rack up 2000 bets by about Xmas. If all continues well I might use 01/01/09 as kick off day.

I'd love to post the selections here but there is seemingly a curse on doing that - might post them from mid december and if they stand up to the forum curse they will stand up to anything!

Thanks again for the valuable input.

Cheers,

Hammers.

[Merriguy]

Hammers,

You seem to be doing well --- congrats. I don't think that there is much I can add; but I would tend to follow Stebbo's advice. There is nothing like the cold hard cash to get you thinking straight.

Otherwise there is the tendency (certainly in my case) to say "Well that just about fits the system. I'll put it in." Disaster. You don't have to start off with a BIG bet. Good luck. :D

[AngryPixie]

After taking in some of your thoughts I reckon a bet size of 1% of max bank level with a strike rate of 30 odd % seems conservative enough to weather the worst of bad runs yet positive enough to enable worthwhile growth. A bank such as this would have reached $14719 today after starting with $1000 39 days ago.
Hammers,

And there in lies the problem with paper testing. Growth like this is very hard to achieve in reality. Other factors come in to play before a lofty goal like that can be reached. I agree with my fellow Punterati - there is no substitute for real money.

You may get something from these

http://www.propun.com.au/racing_forums/showpost.php?p=162311&postcount=46

http://www.propun.com.au/racing_forums/showpost.php?p=162371&postcount=55

[Memsie]

How many bets or what period of time is sufficient to satisfy ones self that the test is reliable?
Hammers you asked the right question and the responses have been very good. But I would be a little bit wary with the fact that your test period is only for the past 39 days unless you have data from before this that franks your system.

Over 39 days you could probably find periods where backing every favourite returns the attractive figures you have quoted so I would recommend that you start off with a small bank.

Your win strike rate is also very high for the volume of bets stated unless you are using the market at very close to jump to determine your selections, this appears not to be case given your longer priced winners. A win strike rate of 31.3% with that volume is equivalent to the strike rate of favourites and in my experience there is no system or person that can select more winners over all races than what the market eventually determines is the favourite. So I would also be a little bit wary of your current high strike rate.

[Hammers]

Thanks for your comments Memsie,

I agree that 39 days isn't long but I wonder whether a method that gives one selection per day for three years is any more reliable than my 1000 odd selections in 40 days as far as a testing period goes - similar volume. It is an intriguing question and I guess only time will tell.

I first toyed with this method last summer and had similar results over a period of about a month - some 700 selections. My laptop was stolen from my car and it wasn't until September I decided to revisit the idea and I was pleasantly surprises to find it still chugging along.

Above all I am realistic. I do not expect 30%+ winners at $4+ average for ever and ever. At 25 or so bets per day I would very happily cop a 5% POT as a $200 bet unit would realise an annual income of $90000 tax free. Given that the excellent betting products available today make it possible to bet in 100% markets or thereabouts I don't see 5% POT as unattainable.

The feedback from the good readers of this forum confirms my belief that 0.5% to 1% of bank on a sound selection method is conservative enough to withstand bad runs yet aggressive enough to encourage growth.

Cheers,

Hammers.

[michaelg]

Hi, Hammers.

You mention a hypothetical method that covers 3 years but with not too many selections.

I would feel more comfortable for a profitable system over a 3 year period because I would assume it would look at all types of races - low class ones to black type which would also include some races where horses are being run in unsuitable races solely to prepare them for a higher race. And also, the system is being bet during the four seasons of the year.

However, not knowing anything about your selection process I cannot really sugget that the time factor would be more important than the number of races to be bet, or vice-versa.

In spite of saying that, if there are multiple selections per race in your method I would be inclined to say that the number of races might then be of more importance than the time factor.

However, if I had a system that performed just as well as yours I think I would have put real money on it by now, even if it was a small amount. And if the strike rate continues then the staking plan can be even more aggressive. Imagine what Maria's result would have been if she had applied a higher percentage of her bank!

Good luck with it.

[Stix]

Hi, Hammers.

You mention a hypothetical method that covers 3 years but with not too many selections.

I would feel more comfortable for a profitable system over a 3 year period because I would assume it would look at all types of races - low class ones to black type which would also include some races where horses are being run in unsuitable races solely to prepare them for a higher race. And also, the system is being bet during the four seasons of the year.

However, not knowing anything about your selection process I cannot really sugget that the time factor would be more important than the number of races to be bet, or vice-versa.

In spite of saying that, if there are multiple selections per race in your method I would be inclined to say that the number of races might then be of more importance than the time factor.

However, if I had a system that performed just as well as yours I think I would have put real money on it by now, even if it was a small amount. And if the strike rate continues then the staking plan can be even more aggressive. Imagine what Maria's result would have been if she had applied a higher percentage of her bank!

Good luck with it.Let's face it's better to lose your money by your own hand than by someone elses.... i.e. following other tipsters. I'd rather put faith in my own abilities than following/depending on someone elses.....

[jfc]

No one here has managed to provide the correct answer.

Which is:

Not only should you be betting this now with Kelly Criterion staking, but you should have started around bet 160.

This question occurs time and time again, and I have answered it time and time again, but for some reason no one seems to be able to grasp the simple concept and retain it.

[stugots]

kelly - full, 1/2, 1/4? which?

& probability of winning %, how does one calculate this magical figure? pulling it out of you know where seems a common enough method...

an example based on hammers numbers would be most welcome jfc

[jfc]

kelly - full, 1/2, 1/4? which?

& probability of winning %, how does one calculate this magical figure? pulling it out of you know where seems a common enough method...

an example based on hammers numbers would be most welcome jfc
Full Kelly. Fractional Kelly is only for those who don't understand what Kelly means.

Hammers' observed strike rate of 31.3% is 305/976

statpages.org/confint.html

Feed that in to the confidence tool to find:

28.35% is the strike rate that you can be confident of exceeding.

Presto.

[stugots]

thanks, so there was a rabbit in the hat after all:)

[Bhagwan]

Another idea for testing , but not as accurate .

Is to have 2 mnths results.

Then run live with small amounts of money for the next 2 mnths.

Cheers.

[stebbo]

Hammers,

do you have a breakdown of the performance (Number, S/R and POT) for selections that start above and below $3.50 ?

Cheers,
Chris.

[Hammers]

Hi Stebbo,

Unfortunately no. I only have the prices for the winners. Are you suggesting some sort of cull or scaled staking idea?

Cheers,

Hammers.

[Memsie]

Hammers, given the volume of your selections you have enough data to divide these selections into various price ranges and calculate the Kelly criteria for each of these. eg <$3.50,$3.50-$5.50, $5.50-$9.00, $9.00-$21.00, >$21.00. But you would need to know the price of every selection not just the winners in order to calculate the Strike Rate (Prob. of winning) for each price range.

What you then would also need to factor in is - what is difference between the best price you can get and the price you used (SP or Home Tote) for each price range. Typically the shorter priced selections do not show as much profit but the Best Fluc for these can be more than 10% better than SP or Home Tote.

[Stix]

Hi Stebbo,

Unfortunately no. I only have the prices for the winners. Are you suggesting some sort of cull or scaled staking idea?

Cheers,

Hammers.(my) Number One Rule: Keep Detailed Betting Records