Overcoming costs questions

[Shakone]

Consider an instrument with costs of one pip/tick/point. Choose a particular time, say noon. In honour of the hare, we flip a coin to decide whether to go long or shot, at market, and we exit at market the same time next day. No stops or targets.

At end of year (say 250 trading days), would you expect to be down 250 pips/ticks + or - some random element? I.e. in the long run, are you going to be down 1 pip multiplied the number of trades?

What do you expect to be down if you enter with a limit order i.e. trying to capture the spread (this is a bit tricky for forex)? How much better or worse? What if you try to enter and exit by capturing the spread?

If you wanted to try and improve on this, would you put in a stop or a target (or both)?

 

[tar]

Consider an instrument with costs of one pip/tick/point. Choose a particular time, say noon. In honour of the hare, we flip a coin to decide whether to go long or shot, at market, and we exit at market the same time next day. No stops or targets.

At end of year (say 250 trading days), would you expect to be down 250 pips/ticks + or - some random element? I.e. in the long run, are you going to be down 1 pip multiplied the number of trades?

What do you expect to be down if you enter with a limit order i.e. trying to capture the spread (this is a bit tricky for forex)? How much better or worse? What if you try to enter and exit by capturing the spread?

If you wanted to try and improve on this, would you put in a stop or a target (or both)?

Technically yes you should be down 250 ticks a year on the long run , assuming no slippage , no black swans and an infinite account to withstand drawdowns . Using limit orders in forex ECN , may help but not that much cuz usually you will get filled when the bid becomes the Ask and the Ask becomes the Bid ( you will pay the spread ) , however using limit orders may help in slippage , anyway this sort of trading "capturing the spread/Marketmaking" is much more suitable with slow markets like STIR , the spread is 1 tick and the daily ATR is just few ticks , you need other traders to trade at your limit orders without prices moving " hit your bid and Ask " , and this can be done in STIR not forex ...

 

[Shakone]

Ok, so in the long run down 250 points perhaps less if we can save by capturing the spread. Agree with you on Forex, but I'm just assuming a general market.

And then if we flip the coin and it is the same direction as yesterday, then there's no point exiting and re-entering, so we can save the cost there. So that's now about 125 points down per year that we expect to be, perhaps less if we can get filled on our limit orders. With no skill, no great entry, no wonderful management of positions, right? Of course we do need to trade fairly small size.

 

[Liquid validity]

Consider an instrument with costs of one pip/tick/point. Choose a particular time, say noon. In honour of the hare, we flip a coin to decide whether to go long or shot, at market, and we exit at market the same time next day. No stops or targets.

At end of year (say 250 trading days), would you expect to be down 250 pips/ticks + or - some random element? I.e. in the long run, are you going to be down 1 pip multiplied the number of trades?

What do you expect to be down if you enter with a limit order i.e. trying to capture the spread (this is a bit tricky for forex)? How much better or worse? What if you try to enter and exit by capturing the spread?

If you wanted to try and improve on this, would you put in a stop or a target (or both)?

Similar to what I do actually, although in this case
I'd guess it wouldn't make much or lose.
Simply as there is no control over losses.
I use static timed session start entry, but exits are anything but.
Thats the only reason it works, cutting losses and running winners.
The major disadvantage is prolonged periods of treading water or drawdown,
which most people wouldn't stomach.
Good discretionary is more efficient in that regard.
The criteria you outlined would be easy enough to backtest anyway.

Capturing the spread is something else entirely, not easy to program at all,
since you are basically entering the quote stuffing arena, post a buy at bid,
to get filled, you need decent size so you don't f**k up the order flow,
limits need constant amendment to match or beat current bid.
Most of all you need an ECN or ECN like platform to do it.
MB, LMAX etc.

As for teh Hare, still p1ssed off about that, but I spose he knew it was inevitable
and didn't care anymore.
If it wasn't for threads like this, not sure I'd bother anymore
So nice one for an interesting post

 

[rawrschach]

Technically yes you should be down 250 ticks a year on the long run ...

this also assumes that there is no skewness, that variances are well behaved etc which is all untrue

 

[wackypete2]

Some ideas to ponder:

How would the time of day affect the outcome? In theory, it shouldn't, but in reality we know there are certain times each day where the market(s) are quiet vs active and moving.

If you were to try to use stops to improve the results I would expect that to work better in a 24 hour market like forex rather than a timed market like NYSE equities due to risk of overnight gaps to blow through your stops.

A low volume, high volatility stock would carry a different risk/reward scenario than a much smoother moving instrument like an index.

Peter

 

[Shakone]

this also assumes that there is no skewness, that variances are well behaved etc which is all untrue

What effect do you think the skew and variances misbehaving will have in the long run on a random direction system?

 

[counter_violent]

Have you considered grids?

Should we start a grids trading thread and see if the collective can crack the problem?

 

[Shakone]

Have you considered grids?

Should we start a grids trading thread and see if the collective can crack the problem?

Yeah, why not?

I do find it interesting, that many will be doing much worse than 125 points down for the year, and will be using all sorts of methods that they think help them.
It's at least worth considering.

 

[rawrschach]

What effect do you think the skew and variances misbehaving will have in the long run on a random direction system?

Literally anything. The full spectrum of losers, break eveners and millionaires.

I'd guess along the mythical 90-10 split of non profitable - profitable.

There are plenty of people making good money safe in the ignorance of the fact that they have no idea that what they're doing isn't better than random entry or is a time bomb.

Survivorship bias etc.

 

[Hakuna Matata]

would also depend on what you were trading (kurtosis)

 

[Shakone]

Literally anything. The full spectrum of losers, break eveners and millionaires.

I'd guess along the mythical 90-10 split of non profitable - profitable.

There are plenty of people making good money safe in the ignorance of the fact that they have no idea that what they're doing isn't better than random entry or is a time bomb.

Survivorship bias etc.

Yes, I think that's a good point. People could possibly make money in the short term with just this or could lose heavily. As I said, 250 points down + some random factor which can be positive or negative, for one year. The random factor may be quite significant for only 250 trades and we may be well ahead or well behind.

As I see it, the system is a losing system in the long run (and over multiple instruments). Do you disagree that it is a losing system in the long run? As you rightly point out there are outliers, but that's not something that will help us.

The system is approximately zero expectation without costs, over 250 trades it is
(-250 +0+ random factor)/250

or put another way, you lose about 1 point per trade. Obvious? This is all in the long run. It is perfectly understandable if you believe that it doesn't converge to a loss of about 1 point per trade, and that the random factor is so large that it dwarfs everything else. That requires some testing.

Every system will have a random factor. We don't want to be slaves to chance, so we have to believe that in the long run the random factor becomes less and less relevant, and the expectation becomes more important. Different systems may be affected differently or take longer to show their true colours. Perhaps some never will.

I don't believe Skewness has as much effect on the results as you think. Although it might do if we start changing the rules for exits or entries. So although these things like variance, skewness and kurtosis have an effect on the random factor, at least some of it is nullified by the way we have set up the system. So can the effects you and Hakuna mentioned be found in time, and are any of them worth more than 1 pip per trade on average?

 

[DionysusToast]

Consider an instrument with costs of one pip/tick/point. Choose a particular time, say noon. In honour of the hare, we flip a coin to decide whether to go long or shot, at market, and we exit at market the same time next day. No stops or targets.

At end of year (say 250 trading days), would you expect to be down 250 pips/ticks + or - some random element? I.e. in the long run, are you going to be down 1 pip multiplied the number of trades?

What do you expect to be down if you enter with a limit order i.e. trying to capture the spread (this is a bit tricky for forex)? How much better or worse? What if you try to enter and exit by capturing the spread?

If you wanted to try and improve on this, would you put in a stop or a target (or both)?

I'd expect you to blow up your account doing this. It's literally an accident waiting to happen.

Of course, the accident could be on your side too.

There is no edge in this approach. In good times, you will lose the spread. You will continue to do this until a black swan takes you out.

Entering with a limit order or stop order is an interesting point but it won't make up the spread consistently. Neither would it see you enter every day. To get filled on a limit, you actually need people to trade against you and to some extent there is a small negative edge there. This is because your bet moves from "I need it to trade up or down" to "I need it to trade down, fill my buy limit and then move up". That is a much harder call to make.

If you don't get a fill (because your coin toss was more right), then you end up not entering the market at all - UNLESS you keep tossing the coin until you get a limit fill. That effectively means you keep tossing, passing on good entries and getting in when you get a chance to step in front of a market moving against you.

 

[Shakone]

I'd expect you to blow up your account doing this. It's literally an accident waiting to happen.

Of course, the accident could be on your side too.

There is no edge in this approach. In good times, you will lose the spread. You will continue to do this until a black swan takes you out.

Entering with a limit order or stop order is an interesting point but it won't make up the spread consistently. Neither would it see you enter every day. To get filled on a limit, you actually need people to trade against you and to some extent there is a small negative edge there. This is because your bet moves from "I need it to trade up or down" to "I need it to trade down, fill my buy limit and then move up". That is a much harder call to make.

If you don't get a fill (because your coin toss was more right), then you end up not entering the market at all - UNLESS you keep tossing the coin until you get a limit fill. That effectively means you keep tossing, passing on good entries and getting in when you get a chance to step in front of a market moving against you.

Yeah I agree, zero edge. So we'd lose at the rate of costs per trade, but may even blow up sooner, depending on size.

On your point about capturing the spread. It's not clear if you're saying you think your losses will go up or down from this? You could miss out on being filled, and that trade might have been a winner. But what is the likelihood, that if you place an order at the bid or ask, that it won't be filled within 24 hours (this may depend on the time you choose as the entry time and the instrument)? Is it outweighed by the missed entry?

 

[rawrschach]

Y...As I said, 250 points down + some random factor which can be positive or negative, for one year. The random factor may be quite significant for only 250 trades and we may be well ahead or well behind.

the point is that the variance or 'random factor' is so large that literally anything could happen.

As I see it, the system is a losing system in the long run (and over multiple instruments). Do you disagree that it is a losing system in the long run? As you rightly point out there are outliers, but that's not something that will help us.

Everything is about outliers. The successful in any competitive field will be the outliers.

The system is approximately zero expectation without costs, over 250 trades it is
(-250 +0+ random factor)/250

You can't realistically calculate expectancy without accounting for variance. When you have a potentially infinite variance any concept of expectancy is useless. Especially when you're trading without some sort of stop (hard or soft).

Every system will have a random factor. We don't want to be slaves to chance, so we have to believe that in the long run the random factor becomes less and less relevant, and the expectation becomes more important.

For this to be true the approach has to have some sort of skewness in it, e.g. some sort of long gamma approach. Flipping a coin leaves you equally open in both directions and there can be no edge.

 

[Shakone]

Rawrschach, I do think considering luck associated with any system is very important, and isn't really discussed, so it's a great point.

Anything can happen with coin tosses. We could get 1 million heads in a row, 10 million, 10 billion etc. The point is that these are extreme outliers, and are single sample paths. Are single sample paths of any use? We need to deal in expectation - on average in the long run, can we make money? Not could we possibly in one eventuality. Pretty much ANY system can give a profit on some outlier sample path, or could blow up. Of what use is that? Saying the successful in any competitive field will be outliers puts in the assumption that these people are just random variables. I don't buy this assumption.

Are you saying the random factor is so large with THIS system or with any trading system? What would you do to decrease that random factor so that in the long run we could get down to the costs and the expectation associated with the system?

We never have infinite variance in practice. That comes about when putting in some maths and assuming certain models with components with infinite variation like Brownian motion. In practice we deal with real data, and can calculate standard deviations which are always finite.

For this to be true the approach has to have some sort of skewness in it, e.g. some sort of long gamma approach. Flipping a coin leaves you equally open in both directions and there can be no edge.

You mean with no stops and no targets, right? What about your statement if we change these?


So far, there's not agreement that we would in the long run lose approx 1 point per trade, there's not agreement about whether using limit orders would reduce that or not, there is - I think - agreement that this system has zero edge, and one person has suggested that we may be able to turn this into a winning system by applying a certain stop, and running the winners.

 

[rawrschach]

We never have infinite variance in practice.

Tell that to the suckers that got filled on sell orders at a penny on major stocks on May 6th.

and one person has suggested that we may be able to turn this into a winning system by applying a certain stop, and running the winners.

And how do you run winners? And how do you account for change in variance and expectancy when your stop takes out trades that would otherwise have been winners? What will your basis be for a stop? Or drawdown on open profit?

I was just passing by to make comment on the stats element but I'm really not into this coin toss thing. Personally I don't think it's worth any time unless you're just doing it as a distraction.

imo the bottom line is that you need to understand how the market works in order to manage your trades. The entry system should be based on that understanding as should the exit system.

 

[Liquid validity]

And how do you run winners? And how do you account for change in variance and expectancy when your stop takes out trades that would otherwise have been winners? What will your basis be for a stop? Or drawdown on open profit?

Trailing stop is the crudest form.
As for the rest - test it.

I was just passing by to make comment on the stats element but I'm really not into this coin toss thing. Personally I don't think it's worth any time unless you're just doing it as a distraction.

Its a good learning tool, or basis for automation.

imo the bottom line is that you need to understand how the market works in order to manage your trades. The entry system should be based on that understanding as should the exit system.

For maximum efficiency, completely agree, only a lunatic would disagree.