Help with sizing using probabilities (I suck)

Is nooby_mcnoob stupid

  • Yes

    Votes: 3 100.0%

  • Yess

    Votes: 0 0.0%
  • Total voters
    3
  • Poll closed .
N

nooby_mcnoob

Say I've figured out that P(S > K) = p_1 and P(S > K+k) = p_2 where k > 0 and p_1 > p_2

Sometimes, k is too large to make any sort of intelligent decision but if I wanted to short somewhere between (K,K+k) when k is large (say some non-trivial fraction of ATR), I am at a loss for how I would size it based on the probabilities. For now, I just size based on the probability that I am wrong based on a linear interpolation between p_1 and p_2.

It seems there is no other way to do it, and I am not some genius quant so none of this makes any sense either I'm sure :)

Related question would be: how do you size based on probabilities?
 
Do you also factor in max account percentage loss per trade (i.e. a losing trade should be no more than a 2% loss for your account)?
 
fan27 said:
Do you also factor in max account percentage loss per trade (i.e. a losing trade should be no more than a 2% loss for your account)?
More...

Hmm good question, could something like this work for sizing? p_loss*E[loss] = 2% of account.

I'm probably over complicating it, but these probabilities have improved my trading big time so I'm trying to figure out how to incorporate into the sizing. The most obvious is the Kelly criterion which I should probably use.

Edit: should mention that right now the sizing is set by the standard volatility and account % formula.
 
Last edited:
Here is an idea. Let's say you will risk between 1.5 to 2.5% of account equity per trade. Next, you come up with a mechanism to determine signal strength based on probabilities of past trades where you score it between 1 and 5. You then use that score to determine where in the 1.5 - 2.5% range you will risk.
 
fan27 said:
Here is an idea. Let's say you will risk between 1.5 to 2.5% of account equity per trade. Next, you come up with a mechanism to determine signal strength based on probabilities of past trades where you score it between 1 and 5. You then use that score to determine where in the 1.5 - 2.5% range you will risk.
More...

That's a decent approach. So if I have P(S > K) = p_1 and P(S > K+k) = p_2 as above, then the interpolated probability would be something like: P(S > K+k_1) = p_1+(p_1-p_2)*k_1/k where k_1 would be my entry.

Some quant somewhere is laughing, I should probably not do this in my head.
 
The calculation of the move probabilities and the incorporation of account $$$ need to be kept separate to not confuse the issue.

The move probs are mechanically the same as confidence interval calculations -- YouTube is your friend.

Sorry so short, but in a big rush today...
 
nooby_mcnoob said:
Say I've figured out that P(S > K) = p_1 and P(S > K+k) = p_2 where k > 0 and p_1 > p_2

Sometimes, k is too large to make any sort of intelligent decision but if I wanted to short somewhere between (K,K+k) when k is large (say some non-trivial fraction of ATR), I am at a loss for how I would size it based on the probabilities. For now, I just size based on the probability that I am wrong based on a linear interpolation between p_1 and p_2.

It seems there is no other way to do it, and I am not some genius quant so none of this makes any sense either I'm sure :)

Related question would be: how do you size based on probabilities?
More...
nooby_mcnoob,

You forgot to put "NO" in your survey.


Why not scale up sizing each time your account doubles? This will keep things simple for you.
 
You must log in or register to reply here.
Back
Top