Selection bias – how much is too much and general MOC discussion

[ScottMcNab]

Nick Radge wrote:

Yes, this would be a Fill or Kill order.

In essence the system is a little different in my opinion, i.e. the buy limit price is not really the buy limit – its a hurdle where the actual but price must be below…semantics is all.

Essentially the system is buying on an opening gap below that hurdle price. If no buy activated, all orders are immediately cancelled.

I like the idea. Would stop indiscriminate fills on a large trend day to the downside.

As stated in the call I used to trade an opening gap system across all Asian futures markets. My research showed that Asian futures tended to ‘over react’ to a down session in the US and therefore tended to rally off that open.

Is it wrong to think that selection bias not a major issue with a system that buys on open only…if there were 40 buys for 20 positions all at the open then the selection should truly be random ? This would mean a system with limit on open order could increase it’s exposure significantly higher than in systems that had buy limit any time through the day

[ScottMcNab]

In other words, could relax entry requirements significantly…wouldn’t matter (in extreme case) if had more buy orders than available positions every day in a limit on open MOC system (LOO FART for J)

[Nick Radge]

Yes. That would be truly random.

But you wouldn’t do that – you would rank the trades using some method and only ever take the top 20 and therefore actually test accurately….

…and you ‘should’ find that results are considerably poorer. Just my 2c…

[ScottMcNab]

I will go away and think about this more. I can’t see at the moment how I would know which stocks would gap lower than the buy limit at the open such that I could apply a ranking …thought it would be same approach as current MOC …use API to place buy limits for multiple stocks before the markets open and wait and see if any drop sufficiently to trigger a buy.

[Nick Radge]

Quote:

.use API to place buy limits for multiple stocks before the markets open and wait and see if any drop sufficiently to trigger a buy.

Scott,
Well, if its a ‘buy on open’ then you are 100% certain to get a fill aren’t you? There is no waiting.

Unless the terminology you’re using is incorrect and you mean ‘fill or kill’.

[TrentRothall]

in the US do they stagger the open like on the ASX ie. open stocks starting with A-F first then G-… etc if so you would get filled in that order if you had a lot of trades on a certain day.

It still confuses me how they can do that, wouldn’t the first batch of stocks opening dictate what will happen later on? If 75% of the first batch gap down 1% then isn’t there a fair chance the last batch will do similar?

[ScottMcNab]

Nick Radge wrote:

Quote:

.use API to place buy limits for multiple stocks before the markets open and wait and see if any drop sufficiently to trigger a buy.

Scott,
Well, if its a ‘buy on open’ then you are 100% certain to get a fill aren’t you? There is no waiting.

Unless the terminology you’re using is incorrect and you mean ‘fill or kill’.

It appears I have used incorrect terminology.. I want to buy the stock on the open if it is below the limit price…using a limit order..limit on open (hence the LOO)

LE = Ref(buysetup,-1) AND (open<=Ref(buylimit,-1) AND NOT OnLastTwoBarsOfDelistedSecurity) ;
LEPrice = open;

[ScottMcNab]

Trent Rothall wrote:

in the US do they stagger the open like on the ASX ie. open stocks starting with A-F first then G-… etc if so you would get filled in that order if you had a lot of trades on a certain day.

IB doesn’t support LOO for ASX Trent…could try and set API to close positions 1 min after open but not sure how complicated that would be for the different groups

[ScottMcNab]

Maybe a different api for each group so can close each 1 min after they open would work

[ScottMcNab]

Julian Cohen wrote:

Is there any reason the position score code will not work on an MR system too? I can’t see why not but just checking

PositionScore = IIf(BuyPrice

[ZachSwann]

I have just had a thought in regards to Kelly criterion and leverage. For example if the result you get out of the Kelly criterion formula is 20% and no leverage. If 4:1 leverage position size 5%.
What everyone elses thoughts?

[SaidBitar]

Kelly equation depends on 2 things winning percentage that is fixed regardless of the leverage and win loss ratio that changes with the leverage
If we neglect the changes of win loss ratio and we assume it is fixed then you need to respect the testing environment if you tested with 10 positions you need to trade 10 positions same is for 20 or 40 positions otherwise the results will vary
If the result is 20% it means the optimal position size is 20% regardless of the leverage and your maximum positions is 20 then you are already at 4:1
If your max positions was 40 then you are at 8:1 which is not feasible so if you trade with 4:1 i.e. 10% you are already at half kelly
So the optimal return is max positions x kelly position size but this will give you volatile equity you can test it and check and this is why i think they half it so lower returns for a smoother equity curve

[ZachSwann]

I thought the reason to half it was if the intraday prices all had lows at the same time on a black swan event. IB would auto liquidate If share were more then 4:1

[ZachSwann]

MOC System on US what time is everyone using to cancel entry orders?

[ScottMcNab]

Used 3.10pm

[Author]

Posts