HI Kirill,

Your exemple is exactly how we go about this issue.

Our models (FX focus) send out signals that are not based on chartpoints, so no price acceleration is expected when a signal is given. Apart from that, when placing a limit order to improve on the price, one has to give the market enough time to do so. Our major models run on hourly data analysis and that is exactly the time we give the limit order to get filled. If it doesn't get filled, the limit order is cancelled and we trade at market.

The two combined (no chart point acceleration and one hour to get limit order filled) has proven to be enough to pay for the few times we do not see a fill and have to go at market at a worse rate.

Of course, as you trade automatically, you must be certain that the platform supports these kind of orders. I have come across a number of platforms that do not accept limit orders (pending orders) when given via Fixapi. Crazy, but true

I hope this is somewhat helpful.

kind regards

Hi Marc, thanks for your input! It strengthens my case.

We trade at a higher rate (5-10 minutes data units). So far we are considering the following options:

* Simple time out before limit order sequence becomes market

* A trigger level when it should become market ( For Ex.: for a buy you place an OCO group Limit Buy and a Stop Buy above for every chunk. If the limit gets filled you place the next pair Limit-Stop. If the stop gets filled you know you got triggered and you continue with market orders until the sequence if complete )

* A mix of both

what do you think ?

NB: everything we use is homegrown and even though in the beginning it was very tough now prototyping things like that has become a fun.

Interesting to see that those same issues are thought with so different time frames. I'm running some algos at 2 sec bars and fighting with limit orders (not get filled) or market orders (not so nice fill price if higher spread) and as those trades are usually less than one minute, even tick or two are quite meaningful (although the number of contracts is very low).

The current model is to run a market orders and not to trade if spread is high as got better results with it as limit orders did not get filled if price was going to the right direction.

Maybe a combination, trade with limit order if high spread, enter a limit order somehow based to bid/ask could improve the results still. (Not so easy to backtest).

Kirill,

We also have a 5 minute model (counter trend) in Fx where we try to get just 1 extra pip on each trade, but to be honest, if you are a market user (not a market maker), that extra pip is often hard to get especially if markets are quiet. So we are programming automated timewindows to trade.

These things are easy to backtest.

Not quite sure how you would backtest over a longer period a mixed approach as you suggest

rgds,

Kirill,

I like the hybrid idea approach. Instead of a 'timeout" approach I would suggest incorporating 'instantaneous liquidity' in to the decision making. If the volume is 'higher than usual' probably you can lock in those extra pips with higher certainty.

A universal solution to this problem does not exist because it's mostly dependent on the particular ECN and especially your permission with this ECN. More information about it could probably help me giving a better advice.

Aleksandar: so far makes no sense because if you place pairs LIMIT-STOP and there is liquidity you gradually pull prices at the LIMIT; timeout approach statistically made no sense so I believe we are going to disregard it completely.

Alex: thanks for your contribution however what I originally asked was not related to any ECN/broker in particular but rather how would one get around the execution bottleneck problem. There is nothing special about the precise moment when we enter of exit a postion so I can not give any qualifiers; I just want to minimize the slippage effect ..

Somebody mentioned combining different venues - it is implied because we use a feed aggregator. And yet it seems that it is virtually impossible to completely get rid of slippage )))

Regards to everyone!

Few limit orders. First (worst) limit order has min size. Second - more than first. And so on. Not sure about right term but we call it "Ladder algo".

Usually algo based on volume trigger. Like VWAP or TWAP. But for fast open/close position algos best way to use a price at the first.

reduce the slippage is more complicated and depending from volatility instrument that we use. Exists two slippage:

* the slippage volatility

* the slippage for sync the teorichal position with real position called slippage time

the only method to reduce slippage it's using order limit if filled without problem

A would like to make a clarification :

* slippage price ( volatility) it shows up with market order

2) slippage time it shows up with limit order

Vitantinio, I would classify the first one as a Liquidity(rather than volatility) slippage which is due to bid/ask spreads.just a thought

Mikhail: thanks !

for me one you say it's no correct completely because the gold future is more liquidity with small bid ask 0.1 but the slippage is too much respect mini SP that have also little bid ask. This considering volatility. The first slippage problem is you use the order, stop, limit or market order but also Tecnology problem, latency and so on

This discussion leads us to the second question of a 'good' setup.. while for exchange traded products it is more straightforward I would like to hear from people who trade FX spot & forwards ...

the slippage forex depends even by the broker with you use. Marker Maker or ECN is most different with slippage. MM even more that ECN

let's speak about serious setup please. At least STP PoP

sry: straight through processing and prime of prime .. as an example ...

Kirill, what liquidity profile is available to you? Any estimate of a typical order size?

Alex: we are limiting ourselves not to have FX positions larget than 5m at the moment because of a potential liquidy trap at the stop. One way to deal with that is to incorporate slippage effects ( at worst ) in the calculation of the position size. Is it a second degree equation if someone is interested I can post it here and formally it resolves the issue. Another way to improve is to sharpen the entry ( because you can hardly sharpen your stop ) and for that we use accumulation/distribution with tickets ranging from as low as 100k to 1m

We define slippage effect as : pips / lots Ex.: 0.00005 / 1 000'000

Kirill, you seem to misunderstood my question. Let me put it in a more simplistic way: how much liquidity is available to you and what is your number in the queue? In other words, how much does your prime broker/bank allow you to take from the market?

In my experience 10m for EURUSD is quite a small amount and I wouldn't care about execution problems at all. Besides that most primes offer ready made execution algos simply in form of special orders, you may want to consult yours.

I used to use a very simple algo which analysed the available liquidity and absorbed it swiping as much as available, or with certain restrictions which prevented me from being considered as a source of toxic order flow. Works like a charm with sizes up to 100M.

Alex: thanks again for your input!

* Liquidity: we trade via a major ECN and I did not understand what you meant by "how much liquidity" - as much as there is !

* EURUSD is a very special case our concern is on crosses ( GBPCAD, AUDNZD etc ... )

3) Utilize built-in algos from the exchange: yes that's a going point which we did not investigate thanks !

Kirill, do you understand how ECN and primes operate? Sorry for asking this, but your answer makes me think that you somehow lack this idea about what ECN participants see in the "Level2". The liquidity you see depends only on the prime broker/bank. Unless you are a liquidity provider you will never see "as much liquidity as there is".

Also, do you want to exploit some peculiarities in crosses pricing? What prevents you from making synthetic crosses? Liquidity there won't be an issue at all. Especially that you say you use an aggregator. Too many vague points about your setup still, unfortunately.

Alex, yes I do understand how they work but honestly I don't like your mentor style. I got in touch with some people who had actionable ideas about what I asked and I guess my question is exhausted thanks to all

found something interesting which exactly pinpoints the subject: http://www.pragmatrading.com/solutions/pragma360-fx