Estimating Ask-Side Depletion for Limit-Order Timing

Summary

The document considers how to estimate when displayed ask depth will be exhausted so a trader can place a bid at the previous ask during an upward price move. The setting is a continuous double auction on a bitcoin futures exchange. A basic regression over recent ask volumes is described as erratic, and the question raises whether event-time models such as autoregressive conditional duration models are needed.

The suggested heuristic compares recent buy volume with current ask depth, submitting when estimated buy volume exceeds the displayed depth. The author reports some success but says the estimate is not very accurate and does not account for irregular trade arrivals. This is a preliminary rule of thumb, not a validated execution strategy; it omits queue position, cancellations, replenishment, latency, and systematic performance evidence.

Key ideas

• The execution problem is to anticipate when displayed ask depth will reach zero.
• A simple estimate subtracts recent buy volume from ask depth.
• The heuristic signals submission when estimated buy volume exceeds displayed depth.
• Trade arrivals are irregular, and the proposed estimate is acknowledged to be inaccurate.

Tags

# Market making: Predicting when to submit a bid at the ask


# Market making: Predicting when to submit a bid at the ask












The exchange I'm trading on lacks a workable stop limit order type (and just deprecated OTO orders), so in order to be first at the old ask with my bid on an up tick, I have to predict the time when the volume at the ask will reach zero.

Does anyone have any pointers for how to go about modelling this? I've tried a simple linear regression on the last 10s of ask volumes but the result is erratic. Do I need to be looking at things like ACD models, or are things a lot simpler than that?

edit: the exchange in question is a continuous double auction, futures exchange trading bitcoin

## Answer by wildbunny (score 1)

https://quant.stackexchange.com/a/43329

I've found some success in using a simple heuristic of the last 10s of buy volume subtracted from the depth at the ask, although this isn't terribly accurate and doesn't capture the fact that trades do not arrive at regular intervals.

Submitting when askDepth - buyVol < 0

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.