Hi there,i'm currently performing a study on intraday cointegration (baskets vs future), the obtained tracking error looks stationnary, dickey fuller/perron tests performed which confirmed the stationnarity. But now i'm facing a new problem, the observed bid/ask spread of the stat baskets are too wide compared to the futures.So then the question would be : how to enter into profitable trades when the basket spread is too wide ?considering limit orders instead of market orders ? but then what about the slippage and the risk to miss the hedge ... ?

1. obviously you need some model of market liquidity for your basket and may be even for for the futures. This will strongly depend how do you execute your basket (order book in case of exchange or other models if you have non-exchange execution components).2. execution slippage components (speed and market depth) are the essential components of your strategy. You basically play with all 3 sub-components: tracking error, basket components market depth, execution speed (fastest tracking basket) to get "working" index arb strategy.

Thanks for your feedback Tu,1. my target is the study of electronic exchanges (order book). What do you mean by a market liquidity model ? what kind of model is it ?2. following the tracking does not looks to me the big deal. But studies/backtests based on components market depth seems to be a nightmare.and how to implement the strat, it's certainly taking ages, no ? i guess that index arb requires high speed access.

the best model is to test your order execution with limit order book simulator assuming your volume does not provide information. It is a nightmare, especially taking into account time synchronization issues between order books for several securities. Probably you could introduce slippage models based on markets volatility but I really doubt there is any meat in index arb without digging in limit orders execution.

by limit orders, you mean trying to grab the flow, or catch the trend. then that is more or less gamblilng on a small period,right ?

QuoteOriginally posted by: zeckpromicby limit orders, you mean trying to grab the flow, or catch the trend. then that is more or less gamblilng on a small period,right ?I don't see why is it gambling because having order book history and transactions history you could EXACTLY backtest the execution of your limit order. The same is true for market orders (especially for huge volumes in high volatility environment).

ok,i tried to google on liquidity models, few papers relating interesting observations (price formation, comovement ...). i'm certainly wrong ... But i don't see where is the benefit from papers trying to modelize markets ???If you have links that can help me , feel free !

there are not too many useful papers. The papers you found probably mostly related to the price predictability, fat tails and market-making vs market-taking (J.P. Bouchaud is a prominent author on this subject). On the contrary you extract price predictability from another source. So your primary concern is to backtest your execution abilities / depth of the market. Try to google The Penn-Lehman Automated Trading Project. Also check web-pages of data providers who provide those (very pricey) limit order book data history.

QuoteOriginally posted by: tu160QuoteOriginally posted by: zeckpromicby limit orders, you mean trying to grab the flow, or catch the trend. then that is more or less gamblilng on a small period,right ?I don't see why is it gambling because having order book history and transactions history you could EXACTLY backtest the execution of your limit order. The same is true for market orders (especially for huge volumes in high volatility environment).not true. having the order book and transaction history is all well and good but it does not include the orders your strategy would place, the mere placing of these orders is going to change the book dynamics in ways you have no way of predicting. if you are backtesting on the less liquid securities this can be a huge problem.

QuoteOriginally posted by: WannaArbnot true. having the order book and transaction history is all well and good but it does not include the orders your strategy would place, the mere placing of these orders is going to change the book dynamics in ways you have no way of predicting. if you are backtesting on the less liquid securities this can be a huge problem.1). See above: "assuming your volume does not provide information" 2). Order information impact is completely different story and taking into account that he trades basket discrepancy which tends to happen during local volatility spikes... it should be reasonably small until you move into really high volume.

usual biblio:- Bouchaud, Almgren, or the nice review by Gatheral to be able to build nice models for liquidity capturing/providing- Almgren, Lehalle, Avellaneda&Stoikov to use such models for trading- Pagès,Laruelle&Lehalle to simultaneously estimate and optimize trading in a dark pool context

Lehalle, which review by Gatheal are you refering to ? Looking at his website i could only find a presentation - Optimal Order Execution. Is this the one ? Also is there a full paper varsion rather than a presentation slides ?

Lehalle,thanks for the biblio. Very interesting, but again i'm afraid that it's very academic and not enough pratical/realistic ...Bouchaud is very descriptive (out of 2 french stocks...) interesting results, just descriptive ... Avallenada/Stoikov is considering markets governed by securities dealers, which is not really valid for european markets for ex.my guess is that all these papers are not very helpfull for building models for real life ... they are only trying to bring a theorical view to practical observations or trying to modelize some unreal markets.

it's not what i'm observing ...