A Trading Riddle re: TP and SL Targets

[garyfritz]

You're (probably) not interested in getting the most wins -- you're interested in maximizing profits per amount risked, so long as it doesn't kill your win% or other stats, yes?

Here's an update to your spreadsheet. It's pretty grubby but it might help you see how this performs. I made several additions:

* Starting in column P, I calculated the W/L for each trade listed in columns M & N. To be conservative, if the SL was hit, I assumed it was a loss, even if the TP was also hit. If neither SL nor TP was hit, I assumed the trade stayed open until the next day.

* I calculated the total profit for the series of trades.

* I also calculated the profit assuming fixed risk%. I said "40 pips = 2% risk," so a 40pip SL traded 1 unit, and a 20pip SL traded 2 units & thus got 2x the net profit.

* I copied the test in column P into the following columns, and swept SL from 20 to 70, PT from 30 to 100.

* I added a win% calculation for each test in P42:BK42.

* I summarized the results in a table in O45:W51, and...

* I plotted the risk-adjusted profits table in a surface chart in A35:G54.

Jem.gif

(Note Excel inverted the SL axis for some reason, so be careful if you compare the plot to the table.)

So from that you can see:

* 40:40 or 40:50 (SL:TP) produce the highest profits.

* However it drops off fairly quickly as SL goes below 40. If you zoom in on the best area:

Jem2.gif

...it looks like 42:45 is about optimal. But this is really much too small a sample to try to force it into that level of detail. There's only one trade with a max negative excursion between 41 and 55, so we have no idea how it actually behaves in that region. It would probably take at least 150-200 trades to get a solid idea of where the best "sweet spot" is.

It's hard to tell without a larger sample, but it might be safest to use a bit larger SL, 45-50 or so, to get away from the dropoff as you get down below 40. Future behavior will almost certainly be different to past behavior, and your actual peak (best profits) in future trades might move around a bit. And the bad-performing areas might move to where the current best peak is, since they're pretty close. It's a good idea to select a position that's in the middle of a big plateau, rather than a sharp peak, especially if the peak is right near a dropoff.

Howzzat?

[Jemook]

Awesome analysis! Gonna have some fun playing around with that, thank you. I agree that adding a few pips to the SL result will help us not get stopped out as often. Also reducing the TP result by a few pips will also allow it to get hit more often.

I have another question now and it relates to baskets. Based on the spreadsheet I did a bit of testing last night and came up with these 'optimal' TP/SL settings for the following currencies based on the past 15 periods:

AUDCHF 40 40
CADCHF 25 25
GBPCHF 60 60
GBPCAD 60 60
GBPAUD 45 45
EURJPY 50 50
EURUSD 40 40
CHFJPY 35 35
EURNZD 65 65
EURGBP 30 30
EURAUD 40 40
EURCAD 35 35

So say I have these pairs in my basket and I want to choose an AdaptiveTP and AdaptiveSL for the basket based on those results. An AdaptiveTP and AdaptiveSL is the number we select to multiply by the number of pairs in the basket.

So if theres 10 pairs and the ATP / ASL is 50 pips then we aim for 500pips TP and 500pips SL.

What would be the best way of choosing a number? I'm thinking just an average may work or perhaps weighted averages to decided the best AdaptiveTP and AdaptiveSL for the basket.

Thoughts?

[garyfritz]

Basing your SL/TP numbers on 15 periods means you don't have many samples in your analysis, right? So the numbers are extremely approximate. But if you're going to trade it as a basket, it doesn't need to be exact. The basket averages everything out.

Averaging the values for the ATP/ASL seems reasonable to me.

[Jemook]

Cool well now that we've gone through all of this and I've learnt a lot I think the best way to work out what to target for baskets is just to keep a journal of my baskets and record the max loss per basket and max gain per basket. This was a fun exercise and I still think there's some validity but a better way would be to fil up the spreadsheet with MaxGainPerBasket and MaxLossPerBasket and work out the best TP/SL for baskets.

Thanks again Gary!

[garyfritz]

No worries. Let me know if you'd like a proper backtest of this beastie.

[shr1k]

this is great I am also trading a basket using a % loss on account as a stop. I am using a rough pip stoploss in the calculation of my trade size. I have been struggling with my stops. Its very hard to backtest any basket strat so most of my testing is forward testing on demo or with tiny lot sizes. the samples are to small to get much useful data from. I have downloaded the spread sheet and will play with it some. Like I said I dont have much real data to use but it may help me some. Thanks.
I am using a MM scheme that uses a group of trades in a series to maintain a positive risk to reward ratio. In my case a higher win rate on the individual trades is important. I can afford an upside down RR ( big stop small TP) on an individual trade as long as the group of trades makes a profit. each trade is small with a small risk.
Its hard to explain to someone that has had never risk more then 2% and let your winners run! That without a Martingale you can be pips negative and dollar positive at the end of the day. This market is not nice and does not follow any real rules For me I do best when I risk a set amount and take my profit in little bites. Any optimization in the bite size VS. the risk size will help me.
I will have a look at the sheet and see if I can come up with better numbers then what I am using now. I will share what I find with a few more details about my entry conditions so it might be helpful to someone else.

[shr1k]

ugg this is a lot more complicated then I thought.
If you dont mind me in the discussion I will share what I am trying to figure out its along the same lines the previous posts.
In my system I am using a set equity loss as a stop that amount does not change until the goal has been met or the loss has been reached.
Based on the assumption that a randomly placed trade long has a 50 50 chance of going 10 pips up or down then it seems that if the stop was 10 and TP was 10 you would have a losing system due to spread. Now if the stop is 20 and the TP is 10 it would seem that your win rate would double but the expectancy of each trade would drop and you would still have a losing system. This is all basic stuff. Now when a trade is placed not randomly but at a point that increases the likely hood of a win it changes things.
My system increases the stop on each group of trades following a winner. The set risk stays the same but the trade size can increase due to account equity being further from the loss cut off. The goal does not change its also a set % gain. this all has the affect of moving the TP closer and the stop further away from the entry. I am trying to figure out an ideal over all risk% and an ideal goal for each set of trades.
The missing piece I have is the real win rate of my entry method at set pip stops and goals. I need to back test my system some more to get a better estimate of my win rate.
Dose anyone have any thoughts on how to dial down my risk and goal settings once I have some win rates figured? I have been using an equity curve sim. and looking at expectancy but its trial and error and only tests one set of settings at a time.