Same results, different trends

[ShaunWhite]

I thought I'd throw this one out there as a bit of a thought experiment...

I took about 3800 results.....and randomised the order in which they might have occurred (all quite possible, ie not just wins first then losers etc). They all obv have the same start and end balance. Plot them, throw on the std Excel linear trend, and project fwd a similar time.

The predictions have a 70% variation. In this case from 550 to about 950. Now I'd expect that using a polymonal trend line or some of the other types but where would you go apart from linear trends to get a more compact set of outcomes? Start & End and draw a line through it seems to be the simple way.?...but then measure how much that deviates along the way to see if it's a 'reasonable' fit?

But perhaps that's just saying that randomising your past results is a good way to get a range of likely outcomes, rather than a single point answer which will always be hard to live up to. It might be a way to alleviate some of the worry of a strategy appearing to underperform; in the case above maybe you just have started out on the blue path and the 'real' trend will be nearer the orange? And that would be fine, but unless you knew that, then you might be tempted to tinker with an otherwise solid strategy that seemed to be slipping vs expected, when it was actually performing within expectations all along.

Screenshot_9.png

I found it quite interesting, maybe some other will. It's getting late but I quite fancy running 1000 iterations and doing a heatmap of the trend predictions. It might lead to a new 'sanity scale' for strategies with a measureable worry%, for any that massively over or under perform into or beyond their red zone.

[eightbo]

Great idea to put this together Shaun definitely interesting. Do you have a function you can easily share for randomising the order of a data set?

P.s. Nice title

[ShaunWhite]

Do you have a function you can easily share for randomising the order of a data set?

Just put a rand() next to each result and sort it.
Then cols to count consecutive ups or downs. And if 'too many' then just f9 to recalc and sort it again.
Then copy/paste the results running total col to a results area being charted.

It's just a thrown together thing, nothing fancy. If it's useful I might tidy it up but I get through 100s of rough ideas for things and various views so I tend not to spend long on them. Remembering what they all are a year later is the hard bit but the best tend to evolve and the rubbish just fades away usually.

I'm not sure there's much mileage in this one once the general vagueness of predictions has been highlighted....again.

[CallumPerry]

This is a much clearer example of what I was trying to say in the other thread the other day! Shaun you have explained it so well.

viewtopic.php?f=36&t=18749

I'm thinking of tracking my live results on top of the Monte Carlo line chart as some sort of confidence indicator for myself...

As Simon said a +ve strategy which is just over 50% ROI is expected at times to have a succession of losing trades that really hurts you mentally but my defence strategy is to plot my live results so I can visually see at any one time that I'm actually still within the expected parameters to make a profit longterm; I think it will be good for self reassurance for any manual trading I do as I try to transition my strategy into semi-automated and then fully-automated.

The only difference being I thought plotting my live progress would be a good way to go about it, your idea of creating a heat map is so much more elegant. Please let me know how you get on with this! Very happy to see you have this working as it is something I am going to be using from now on too!