Devil in the detail

[kinglouie]

A relation of mine who just happens to have a very well paid job within the IT industry told me a long time ago that "Computers don't make mistakes, people do" inferring that no matter what you think a computer or software may have done its actually human input/error at some point that has caused the issue in the first place.

I have always remembered that and it's something I use to remind myself that if my trading bots, models, rules etc are not delivering the desired outcome then its not down to the computer or software its down to the instructions/rules etc I have created.

Anyway, I have been testing an in play theory for a few weeks and decided to trial it yesterday in realtime to very small money as I thought I was close to getting it right

Turns out, I was close but not as close as I need to be.

Screenshot 2019-04-22 at 09.05.41.png

On looking back through the logs of the 3 races that gave me a RED outcome, it evidences that my timings were not quite right if a particular set of circumstances occur and I have now altered the rule accordingly along with another error I noticed I had made.

Just thought I would share it as I know how frustrating trading can be and even when you think your close to getting it right for a particular market or sport, you can always use the detail to find out why you're not or what needs to be changed.

I know lots of people who have sound theories that are fairly close to the mark, but fairly close in this environment is often not close enough but often the information that is available can point you in the right direction.

Regards

KL

[MemphisFlash]

your making a mistake thinking you will get a positive green every time you use a particular bot.
YOU WON'T
No matter what you do there will ALWAYS be the negative values somewhere along the line.
NO BOT IS 100% RELIABLE.
if someone says it is they are lying.

[kinglouie]

I didn't say I or anyone else will get a positive green every time, but I did say that you could use the detail to do your own analysis of your bot, rule, system etc to make any adaptations you need to so that you can get it as close as you can to delivering the best outcomes it can when being used to trade.

I am under no illusions that a bot can be 100% accurate all the time, there are too many things that can happen in sport for that to be the case in my view, its far to dynamic and complexed to successfully cover every outcome/eventuality.

[LinusP]

I didn't say I or anyone else will get a positive green every time, but I did say that you could use the detail to do your own analysis of your bot, rule, system etc to make any adaptations you need to so that you can get it as close as you can to delivering the best outcomes it can when being used to trade.

I am under no illusions that a bot can be 100% accurate all the time, there are too many things that can happen in sport for that to be the case in my view, its far to dynamic and complexed to successfully cover every outcome/eventuality.

Hate to say it but Memphis has a point, you also need to confirm that the said mistake wasn’t also giving you all your profit.

[kinglouie]

Not sure I disagreed with anyone in respect of any point they my have made and the error/s had no bearing on the outcome of other races, as I said, I had overlooked a specific set of circumstances.

[ruthlessimon]

On looking back through the logs of the 3 races that gave me a RED outcome, it evidences that my timings were not quite right if a particular set of circumstances occur and I have now altered the rule accordingly along with another error I noticed I had made.

I try to do similar analysis; but I always want it supplemented with long term data.

i.e. The "refinement" arises from a single market (video) - to be tested on a yr's sample of markets (trade sim).

"I lost & there were 12 horses in that race - does the number of horses impact my strategy?" etc

[eightbo]

Is it wise to assume the "particular set of circumstances" experienced today will always play out in the same way as these 3 specific markets in the future?

Please post your P&L at the end of the week if you don't mind thank you

[kinglouie]

Two out of the three were relevant to timings in relation to the race distance (12F Flat and 28F NH).

Interestingly, I make no claims other than to state that reviewing the data can often help people potentially understand why trades are Red instead of Green and using that to adjust/revise a bot, rule or strategy can be beneficial.

I not sure posting my P&L is relevant on that basis or is having to defend claims I have not made in relation to a bot, rule or strategy being able to provide a green figure on every trade.

Too much negative keyboard rattling for me.

Regards

KL

[Vorogue18]

Perhaps I am missing something but to me KL's initial post reads as a 'hang in there, I share your frustration, you maybe close like me so just tweek the detail, don't throw it away". I took a lot from that so thanks KL. No boasting or claims of perfectionism from my angle....

[ShaunWhite]

Perhaps I am missing something but to me KL's initial post reads as a 'hang in there, I share your frustration, you maybe close like me so just tweek the detail, don't throw it away". I took a lot from that so thanks KL. No boasting or claims of perfectionism from my angle....

+1
Give the guy a break ffs

btw KL, adding 10p to 17 winners is sometimes easier than adding 50p to 3 losers. Most advice you'll see says look at the losers first but that's illogcal imo. Making a big difference to a small set of results often isn't as effective as making a small difference to a large set of results. Afterall you're doing automation, automation is all about scale and fixing a few outliers isn't always the answer.

It also the case that you can sometimes make a bot more profitable by making it LESS profitable per race. If you reduce it's individual race effectiveness by 10%, but that means it trades twice as many scenarios, then that's potentially better than a strategy that makes bigger greens but only does so a few times a day.

[ruthlessimon]

Give the guy a break ffs

btw KL, adding 10p to 17 winners is sometimes easier than adding 50p to 3 losers. Most advice you'll see says look at the losers first but that's illogcal imo. Making a big difference to a small set of results often isn't as effective as making a small difference to a large set of results. Afterall you're doing automation, automation is all about scale and fixing a few outliers isn't always the answer.

It also the case that you can sometimes make a bot more profitable by making it LESS profitable per race. If you reduce it's individual race effectiveness by 10%, but that means it trades twice as many scenarios, then that's better than a strategy that makes 100% greens but only does so 3 times a day.

Well it's good he hasn't been given a break; cos your post is f4ckin' solid

That's golden that one, really like it

[ShaunWhite]

That's golden that one, really like it

ty ty. The only important thing is the bottom line. I don't especially give a rats ass about pl on individual races or the overall strike rate.
Tweak it, run it, log the results, is it better or worse?...repeat. Ideally keeping the variables to an absolute minimum so that the number of tweak combos don't get out of hand.

[eightbo]

I not sure posting my P&L is relevant on that basis or is having to defend claims I have not made in relation to a bot, rule or strategy being able to provide a green figure on every trade.

You posted the daily P&L so thought ya might care to share smth longer term. If not no worries.

[kinglouie]

Its good to reflect on posters comments and true to society today, it is evident that there are those who read what they want things to say and those who read what things actually say.

Whilst not directly linked, some of you make take something from this clip in relation to trading, some of you may not, in my view its worth the watch, enjoy.

https://www.youtube.com/watch?v=MzS5V5-0VsA

[eightbo]

Ackoff was a wise individual.

People view the world through their own filters. Assuming you know how others are interpreting information however is just another example of this.