Trading What I see !?

[goat68]

Just added up lifetime account stats,

You should be able to swerve PC completely if you can stay above 20% comm generated. And ideally aim for 40% to make PC2 bearable.

Would be nice, current greyhound bot is only 6% comm generated.
20+% are surely value betting strategies?
anyway got a good runway to work that out...

[ShaunWhite]

20+% are surely value betting strategies?

They're all value betting strategies

I get what you mean but that's not necessarily the case. I hedge (for a sound reason) and generate about 30% ish.

[goat68]

So here's my trading long term PnL, i'm hoping i've just hit the double bottom!
I'm actually feeling very confident about 2022, I reckon I can wipe my long term loss (-£538) by the end of March:

Screenshot 2021-12-31 at 18.07.18.png

[goat68]

So being running some backtests on some pure TA strategies, after having some thoughts from the Supply & Demand discussions, but definitely think pure TA is not for me, they always seem to end up net zero.
Eg.This was looking good to start with, but then just turned on it's head, which seems to be a common theme with pure TA:
Week 1 TOTAL : 92.34 for 7280 bets
Week 2 TOTAL : 101.99 for 8597 bets
Week 3 TOTAL : -276.86 for 9773 bets
Week 4 TOTAL : -194.97 for 9032 bets

[goat68]

This is the best i've come up with for pure TA on uk dogs:

Total profit = 243.03
Turn Over % = 0.49 %
Commission = 119.83 (49.31 %)

Strike Rate = 41 %
Av Win:Loss Ratio = 1.52
Highest Hi = 328.21
Lowest Lo = -33.33
Biggest Drawdown = -157.37
Longest Drawdown = 19 days
Longest Time to NewHi = 37 days
Longest win streak = 10
Longest lose streak = 15

Don't particularly like the fact the biggest drawdown -157 is a good proportion of the total profit 243

Would be a good commission generator though if it does stay 0/+ve

[ShaunWhite]

Don't particularly like the fact the biggest drawdown -157 is a good proportion of the total profit 243

Try to think about other ways to measure the PnL volatility rather than drawdown. £-157 out of £243 looks bad, but would it seem that way at the end of the year if it's £-147 vs £2,430 or vs £24,300 ? Or what about £-147 as an initial run when the total is zero?

I prefer to look at the likihood of a set of results occuring, or #std deviations from short/med and long term trends. "Drawdown" is only relative to something else so you need to define that something else to measure it against other than however much you'd made to date which is arbitary and dynamic.

[goat68]

Don't particularly like the fact the biggest drawdown -157 is a good proportion of the total profit 243

Try to think about other ways to measure the PnL volatility rather than drawdown. £-157 out of £243 looks bad, but would it seem that way at the end of the year if it's £-147 vs £2,430 or vs £24,300 ? Or what about £-147 as an initial run when the total is zero?

I prefer to look at the likihood of a set of results occuring, or #std deviations from short/med and long term trends. "Drawdown" is only relative to something else so you need to define that something else to measure it against other than however much you'd made to date which is arbitary and dynamic.

at least it has the Comm/profit ratio you mentioned, 49%

[goat68]

From watching the activity and volume on some races, i can't see how you can gain enough pure technical edge to be profitable. The real price movements are based on market opinions and there's not enough edge to detect those in a pure technical manner. Imho !

The backtests i have run, if at all profitable were i suspect random variation...

[ShaunWhite]

From watching the activity and volume on some races, i can't see how you can gain enough pure technical edge to be profitable. The real price movements are based on market opinions and there's not enough edge to detect those in a pure technical manner. Imho !

TA is a broad church, from regular order flow trading, to hft betting on certain setups. Basically its everything that doesn't involve knowing about specific dogs.

Are you tending to bet and hold (to sp or to run) or having multiple bets possibly swapping sides several times, similar to how you do it manually?

[goat68]

From watching the activity and volume on some races, i can't see how you can gain enough pure technical edge to be profitable. The real price movements are based on market opinions and there's not enough edge to detect those in a pure technical manner. Imho !

TA is a broad church, from regular order flow trading, to hft betting on certain setups. Basically its everything that doesn't involve knowing about specific dogs.

Are you tending to bet and hold (to sp or to run) or having multiple bets possibly swapping sides several times, similar to how you do it manually?

multiple bets either way, although I limited in my tests to between +4 to -4 on each selection, ie.limiting back or lay liability to 4 bets either side

[ShaunWhite]

multiple bets either way, although I limited in my tests to between +4 to -4 on each selection, ie.limiting back or lay liability to 4 bets either side

So each bet is 1/4 of your max....I just go max of any remaining unused liability, and cancel any unmatched amounts when my signal is lost obv. I didn't find any advantage from splitting up the stake into smaller amounts. In fact it was worse because you don't end up with much on if you only bet once or twice on one side.

That's an easy backtest to check that one of course if you've got a max bet size parameter as well as a max liability parameter, just make them same. Whether '4' or any other number is best is knowable so needn't be a guess if that's what it was.

To check, if you had 4 bets one one side, would your lay bet stake now be double the usual amount? Eg if you're at +4 then you have -8 available to get to the -4 max the other way. Would that now permit 8 lay bets, or 4 lay bets before you got to -4?

[goat68]

Found a new dog model strategy, looks good, not the throughput of the first one, but every bit adds up:

Screenshot 2022-01-02 at 12.20.13.png

[ShaunWhite]

A date on the x axis would be handy, I was wondering when that kink happened about 1/3 of the way along that halved (?) your RoI. Did this one stand up to in & out of sample testing to make sure it's not backfitted?

Also do you get what I was saying about allowing 4 bets? It seems to be reducing the amount you're turning over and there's no reason the 1st bet needn't be any better or worse than the 4th.

[goat68]

A date on the x axis would be handy, I was wondering when that kink happened about 1/3 of the way along that halved (?) your RoI. Did this one stand up to in & out of sample testing to make sure it's not backfitted?

Also do you get what I was saying about allowing 4 bets? It seems to be reducing the amount you're turning over and there's no reason the 1st bet needn't be any better or worse than the 4th.

Yes to in and out samples
The kink is actually due to the spike up 1/4 way, which was a race that started 25sec early and was not hedged, bet won!

Need to think about the 4bet thing...
The 4 bets are at different times..

[ShaunWhite]

The kink is actually due to the spike up 1/4 way, which was a race that started 25sec early and was not hedged, bet won!

? This is the one I was referring to...

Screenshot_2.jpg