Trading What I see !?

[Emmson]

So darkest hour really does come before dawn, only seems a few weeks ago when Goat was at his nadir asking whether to give up. Nice work I hope it lasts, a lot of perseverance on display in this thread.

[Kai]

So darkest hour really does come before dawn, only seems a few weeks ago when Goat was at his nadir asking whether to give up.

Indeed, I like this one from Tony Robbins.

[ShaunWhite]

So darkest hour really does come before dawn,

That's what people say who get to see the dawn. And strictly speaking it's not even factually correct. I think it belongs in the same drawer as those Brad Goodman phrases that must be true because they rhyme

[goat68]

So I thought I would spend some time looking ahead at potential extra Betfair charges, and was looking at the "transaction charge", and can't understand Betfair's example!
Can anyone understand where the £2.10 commission offset value comes from in the Betfair example below? I can't see how you get £2.10 from the £18.50 generated commission?! They state they will offset any transaction charge by the commission generated. They don't however state over what period the commission generated is? is it the day?, or the hour of the charge? Either way can't see how you get £2.10 below...??

Example::
You are an API user who pays (commission + implied commission) ÷ 2 of £18.50 throughout a given day.
However, between the hours of 14:00:00 and 14:59:59 you placed 4800 bets and recorded 3400 failed transactions, between the hours of 15:00:00 and 15:59:59 you placed 2250 bets and recorded 4200 failed transactions, but in other hours you recorded less than 5000 transactions. Your base transaction charge is (4800+3400-5000) X £0.002 + (2250+4200-5000) X £0.002 = £9.30.
Therefore you will be billed an additional £9.30-£2.10 = £7.20 on the following day.

[goat68]

The closest theory I can get is they pro rata the day's generated commission by the number of hours you traded transactions in compared to the number of hours that exceed 5000, and in this example if that was 18 hours (maybe!) in which they generated £18.50, then 2 hours pro-rata from that is (2/18)*£18.50 = £2.06

[goat68]

Someone in betfair is reading my thread!, as they have literally just fixed the doc ...(https://www.betfair.com/aboutUs/Betfair.Charges/), the £18.50 has been changed to £2.10, that makes a lot more sense now

Example::
You are a Betfair Exchange user who pays commission generated of £2.10 throughout a given day.
However, between the hours of 14:00:00 and 14:59:59 you placed 4800 bets and recorded 3400 failed transactions, between the hours of 15:00:00 and 15:59:59 you placed 2250 bets and recorded 4200 failed transactions, but in other hours you recorded less than 5000 transactions. Your base transaction charge is (4800+3400-5000) X £0.002 + (2250+4200-5000) X £0.002 = £9.30.
Therefore you will be billed an additional £9.30-£2.10 = £7.20 on the following day.

[andy28]

Thought I would ask this here as it is about sample sizes.

I have a value betting strategy I am working on, it's looked at in total 25000+ runners of which it has placed 4100+ bets (can select more than 1 runner per race) and has won 605 races. This is all based off my paper bets.

So my question is which is the more important for a good sample size, the runners, selections or the winning bets? I am thinking that 605 winners is nowhere near enough to get a true indication but then I think 25000 runners is a decent size. It's a low strike rate as I am only winning at15% with a yield of 28% which looks good but I tried it in practice mode and instantly took a nose dive but has recovered a little bit. But before I go live are these samples big enough or do I need more?

[goat68]

Thought I would ask this here as it is about sample sizes.

I have a value betting strategy I am working on, it's looked at in total 25000+ runners of which it has placed 4100+ bets (can select more than 1 runner per race) and has won 605 races. This is all based off my paper bets.

So my question is which is the more important for a good sample size, the runners, selections or the winning bets? I am thinking that 605 winners is nowhere near enough to get a true indication but then I think 25000 runners is a decent size. It's a low strike rate as I am only winning at15% with a yield of 28% which looks good but I tried it in practice mode and instantly took a nose dive but has recovered a little bit. But before I go live are these samples big enough or do I need more?

I evaluate these things a bit by my own feel and judgement, based on my experience, others are a lot more precise and analytical.
I would say having two independent samples is important. Also having a reasonable amount of analysis, so you can make judgement say on your live test, as to whether what you see is within the realms of the expected. So things like expected drawdowns, win%, selection rate,... you can then tell how often in my samples did I get such a nose dive..?

[ShaunWhite]

with a yield of 28% which looks good but I tried it in practice mode and instantly took a nose dive but has recovered a little bit. But before I go live are these samples big enough or do I need more?

How are you measuring 'yield'.

28% is about 10x more than I'd expect so if it took a nosedive is it just reverting to something more realistic? People often focus on the losses, drawdowns and downturns without considering if the gains were realistic in the first place.

When you said practice mode was disappointing, did your analysis/paper trading make assumptions about obtaining something unrealistic like high or low process, or was it something more doable like betting at sp? Ie how far from the ideal is your execution managing.

Sample size: don't get too hung up on the number of selections etc. The thing that's important is the time period it covers. It should at least be one full season of the type of markets you're interested in eg flat, jumps or a whole football season. Or if it's more generic then probably a full calendar year.

And then consider what it would be like to live with the strategy in real-time, you'd really hope to show some sort of gains each month because even a big annual profit might be too hard to live with or trust if 9/12 months show losses.

[andy28]

Thanks Guys appreciate your thoughts

Shaune I used practice mode as I wanted to see if I got the SP at the odds I thought, ie if I wanted to get between 8-12 what did I need to use as last traded price. Also when races run behind time how much does it effect them, and it can vary a lot. I got odds of between 6.78 and 13, however over 90% were within the range

The yield I just divided the profit by the bets so was 1150/4100 or there abouts. This is value betting and not trading so I thought 28% is above average but in all honesty I don't know what it should be, I was working on 20%

I entered the results into an online Monte Carlo Sim and it did say Max DD would be around 179 so losing 40 straight is well with in that range. Also when I look at the MC Sim it was below 0 for around 217 bets.

I did think about seasonal impact as my paper tests are for just 3 months. I did randomly select past results going back a year (2 from each month) and it won 7 of the 9 months. Maybe I need to get a full year to truly see what part seasons play

[ShaunWhite]

The yield I just divided the profit by the bets so was 1150/4100 or there abouts.

Better to use profit/total amount staked which is what most people do. Then you can compare your perfomance more easily if/when you alter your stake. Frankly I think you'll be very lucky to find 2-3% after commission, 5% at a push but that would be seen as very decent.

But the massive elephant in the room is how deep your pockets need to be to bet 10/1 shots at SP. Like you say 40 straight losses won't be unusual at all and you need to be braced for 80 or even 180 at a decent stake.

The good news though is finding 4100 selection in 3 months, that's great because Profit is Turnover * Edge, so when a good edge is a single figure %age we need to turn over as much cash as possible.

[andy28]

The yield I just divided the profit by the bets so was 1150/4100 or there abouts.

Better to use profit/total amount staked which is what most people do. Then you can compare your perfomance more easily if/when you alter your stake. Frankly I think you'll be very lucky to find 2-3% after commission, 5% at a push but that would be seen as very decent.

But the massive elephant in the room is how deep your pockets need to be to bet 10/1 shots at SP. Like you say 40 straight losses won't be unusual at all and you need to be braced for 80 or even 180 at a decent stake.

The good news though is finding 4100 selection in 3 months, that's great because Profit is Turnover * Edge, so when a good edge is a single figure %age we need to turn over as much cash as possible.

Well worth going further back, I looked at the 3 months prior to my paper test and each month showed a loss, albeit small it is still 3 losing months. So now showing a profit of 10% for 6 months. Looks to me the 3 months I looked at was a honey hole that won't last. Still need to do the full year but I think it is safe to say it is heading towards the scrap heap.

Lesson learned is there are no short cuts

[goat68]

So a nice result for March, dogs exceeded expectations, horses had a bit of a drawdown mid-month and were a bit weak.

Screenshot 2022-04-02 at 21.28.23.png

[The Silk Run]

So a nice result for March, dogs exceeded expectations, horses had a bit of a drawdown mid-month and were a bit weak.
Screenshot 2022-04-02 at 21.28.23.png

Kool. STICK with the GH's G. Well done ...

[The Silk Run]

So a nice result for March, dogs exceeded expectations, horses had a bit of a drawdown mid-month and were a bit weak.
Screenshot 2022-04-02 at 21.28.23.png

Kool. STICK with the GH's G .Very well done .....