How long does it take to judge whether a system is working?
-
- Posts: 124
- Joined: Mon Dec 11, 2017 6:26 pm
Oh, that was tempting fate... Right after I posted that last message I had another losing trade.
-
- Posts: 124
- Joined: Mon Dec 11, 2017 6:26 pm
Day 5:
I got off to a good start today then hit a really rough patch before gradually recovering. I did 17 trades today bringing the total trades up to 73.
Today's strike rate: 71% (12/17)
Overall strike rate: 76% (56/73)
Today's growth: £12.98 to £13.28 = 2.3% growth of bank.
Overall growth: £10.00 to £13.28 = 32.8% growth of bank in 5 trading days.
Today was my lowest strike rate so far and also my lowest profit growth.
I got off to a good start today then hit a really rough patch before gradually recovering. I did 17 trades today bringing the total trades up to 73.
Today's strike rate: 71% (12/17)
Overall strike rate: 76% (56/73)
Today's growth: £12.98 to £13.28 = 2.3% growth of bank.
Overall growth: £10.00 to £13.28 = 32.8% growth of bank in 5 trading days.
Today was my lowest strike rate so far and also my lowest profit growth.
You do not have the required permissions to view the files attached to this post.
- ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
Might also be worth looking at you return on investment figure (ROI) rather than just 'growth'. pl/stake
Those charts are like fractals sometimes, zoom out 10000 races and you'll see that's a small bit of a big graph that looks exactly the same. Hard part is figuring which bit.
Those charts are like fractals sometimes, zoom out 10000 races and you'll see that's a small bit of a big graph that looks exactly the same. Hard part is figuring which bit.
- ruthlessimon
- Posts: 2094
- Joined: Wed Mar 23, 2016 3:54 pm
& this is what makes trading so difficult!
I thought Peter's shades of grey video was brilliant - although I think labeling it as "psychological issue" was slightly misleading. Knowing the strategy needs to be tested with a slightly different variation - that's not psychology - that's pure trading wisdom
Frog if you can I recommend doing it; just tweak some variables & see how things change.
If I hadn't - I'd thrown this idea out on the first trial. But for some reason the "mid-variations" are outperfroming (I aint got a clue why btw ):
-
- Posts: 124
- Joined: Mon Dec 11, 2017 6:26 pm
Yes, I appreciate the point. I'll not be pursuing this experiment to anything close to 10,000 races though - long before then I hope to know if it's a success or a failure. However... I might do a daily graph showing just the trades for each day, in addition to my main graph, as that might tell me something. All the trades in any one day are to the same stake as I only adjust the stake settings at the end of each day.ShaunWhite wrote: ↑Thu Apr 26, 2018 8:34 pmMight also be worth looking at you return on investment figure (ROI) rather than just 'growth'. pl/stake
Those charts are like fractals sometimes, zoom out 10000 races and you'll see that's a small bit of a big graph that looks exactly the same. Hard part is figuring which bit.
-
- Posts: 124
- Joined: Mon Dec 11, 2017 6:26 pm
Yes, I did make some small adjustments today after the losses. I saw, in the case of two of them today, that if I'd got out on the Green side for a lower profit than I'd been seeking then I'd not have seen those two go red. They're not drastic and don't alter the basic principles of the strategy but I've fiddled about with where my closure points (on the green side - the red side stays the same) are and I'll be curious to see how it plays out tomorrow.ruthlessimon wrote: ↑Thu Apr 26, 2018 8:53 pm
Frog if you can I recommend doing it; just tweak some variables & see how things change.
If I hadn't - I'd thrown this idea out on the first trial. But for some reason the "mid-variations" are outperfroming (I aint got a clue why btw ):
It obviously not easy to calculate what I would have ended up on if I'd tweaked before today's races started but it would certainly have been more as my strike rate would have been 82% instead of 71%. Three of my losing trades would have remained losers at the same amount as before whilst the other two losing trades would have been small positives. Some of my wins would have been reduced in size though so it's a tricky balancing act.
Anyway, tomorrow is another day.
-
- Posts: 124
- Joined: Mon Dec 11, 2017 6:26 pm
For anyone who is curious here is how my graph would look if I'd been using even stakes (based on an initial £100 bank just because it's easier to deal with percentages using £100 than using my real £10 bank).
I'll maintain this version of the graph alongside my other one.
I'll maintain this version of the graph alongside my other one.
You do not have the required permissions to view the files attached to this post.
-
- Posts: 124
- Joined: Mon Dec 11, 2017 6:26 pm
Or maybe it makes more sense to show it this way?:
You do not have the required permissions to view the files attached to this post.
- ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
Without knowing what the strategy is and the race selection criteria, it's just a line.FrogThimble wrote: ↑Fri Apr 27, 2018 12:01 amFor anyone who is curious here is how my graph would look if I'd been using even stakes.
My comment about 10000 races was obviously tongue in cheek but nonetheless true.
-
- Posts: 124
- Joined: Mon Dec 11, 2017 6:26 pm
Race selection is based upon whether the price of the Favourite (which I will try to trade in-running) at 10 seconds before the off is:ShaunWhite wrote: ↑Fri Apr 27, 2018 12:50 am
Without knowing what the strategy is and the race selection criteria, it's just a line.
My comment about 10000 races was obviously tongue in cheek but nonetheless true.
1) Within a certain price range
2) A certain number of points below the price of the 2nd Favourite
I'm sure you'll understand that I don't want to say exactly what the figures are for those conditions but, if both of those conditions are met then I trade the race. If either one or both conditions are not met then I sit the race out. If I play then I try to trade for about 5% of what I'd win if the horse were to win the race. If the horse never reaches the point where I can get out for that profit then I have a safety servant to take me out for a Red.
However... I'm introducing a 3rd condition today about race length.
This is a graph of the percentage growth of my bank at level stakes. The blue line represents all of the races of 12 furlongs or longer that I've traded (almost all jumps, of course). The orange line represents all of the races of under 12 furlongs that I've traded (all flat races).
I think the difference is quite stark. Of course I don't know if I'm looking at a Long v Short or a Jumps v Flat difference but I'm going to treat it as Long v Short. These shorter races are maybe not worth using my strategy on after all.
You do not have the required permissions to view the files attached to this post.
Last edited by FrogThimble on Fri Apr 27, 2018 5:04 pm, edited 1 time in total.
- ruthlessimon
- Posts: 2094
- Joined: Wed Mar 23, 2016 3:54 pm
Now Peter would be a better person to ask, (because I am super new (typo there ) with this type of analysis), but I think seeing divergence like that (if both strategies were deployed simultaneously) should breed confidence that your original strategy is a real edge - & if you know what/why the edge is, then your onto a winner - literally
The question is what's causing the divergence
P.S. I think the divergence is good in this case, as you've added a completely new variable. A minor tweaked variable could suggest randomness - but again Peter is the person to ask about that
The question is what's causing the divergence
P.S. I think the divergence is good in this case, as you've added a completely new variable. A minor tweaked variable could suggest randomness - but again Peter is the person to ask about that
-
- Posts: 124
- Joined: Mon Dec 11, 2017 6:26 pm
My opinion at the moment is that the divergence is caused by the lead changing hands more frequently and earlier in the short sprint races. As soon as the horse I am trading loses Favouritism then I go into the Red. In the longer races the original Favourites seem to retain favouritism for a longer proportion of the race even if they are going to lose - so more chance to grab my profit. On the other hand the danger in the longer races is that if my horse falls at a fence before I've done the 2nd half of the trade then I'm looking at a bigger loss - but I can't do much about that really.
- ruthlessimon
- Posts: 2094
- Joined: Wed Mar 23, 2016 3:54 pm
That strikes me as pure knowledge, & an interesting bias if true
-
- Posts: 124
- Joined: Mon Dec 11, 2017 6:26 pm
No, not knowledge so much as observation. I know nothing at all about horses but I'm coming at this from a mathematical bent. I've been watching the price movements carefully and it seems to me that a lot of really wild stuff happens to the prices in the early stages of the short races whilst the early stages of the longer races are more stable.ruthlessimon wrote: ↑Fri Apr 27, 2018 5:18 pmThat strikes me as pure knowledge, & an interesting bias if true
- ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
The course also has a significant effect on IP strategies.
Whether you can prove much by backtesting is a hotly debated topic but you can certainly use it to look for the likely effect of course, runners, surfaces, distances, seasons etc. Or just quickly see if a strategy is a non runner.
These examples aren't related what you're doing, but are from strategies I tested on the last 50,000 races run in the UK. A few thousand more races and it might be a decent sample size Actually the data goes back 100,000 races, 1 million horses over 10 yrs, but it just takes too long to run my iterations.
Maybe you should have a play with the data, if you're a maths guy you might find something.
BUT you're much more likely to find something if you start collecting your own data to zero in on the area you're interested, ideally something which isn't in the public data.
Example BF data...
Whether you can prove much by backtesting is a hotly debated topic but you can certainly use it to look for the likely effect of course, runners, surfaces, distances, seasons etc. Or just quickly see if a strategy is a non runner.
These examples aren't related what you're doing, but are from strategies I tested on the last 50,000 races run in the UK. A few thousand more races and it might be a decent sample size Actually the data goes back 100,000 races, 1 million horses over 10 yrs, but it just takes too long to run my iterations.
Maybe you should have a play with the data, if you're a maths guy you might find something.
BUT you're much more likely to find something if you start collecting your own data to zero in on the area you're interested, ideally something which isn't in the public data.
Example BF data...
You do not have the required permissions to view the files attached to this post.