Real Trade Price Execution (Order Fill) Analysis + Discussion On Efficient

[Trendnomics]

Hey guys/girls,

I've been analysing my trade price executions (order fills), for the past three and a half years of trading.

I analysed my buys and sells separately, as follows:

• Obtained the High and Low price for the equity on the day of trade execution - I then calculated the Median price accordingly;
• Calculated the percentage deviation of my Executed price from the Median price, bounded by the High/Low price - I ensured that each calculated result's symbol (+/-) reflected the correct outcome (i.e. + = value add / - = value loss) - for example (Executed buy price = 4.51), (Median price = 4.675), (Low price = 4.48) -> (Deviation from Median Price = +84.62%);
• Calculated the average result for all the Deviation from Median Price results;
• Calculated the number of Executed prices that were value adders (i.e. above/below median price - as applicable to buy/sell).

The following results were obtained for +500 real executed trades - note: trading commissions were not included in this analysis (purely price based), ASX300 universe and most trades were executed between 12pm-3pm:

For Buys:

• Average Deviation from Median Price = -1.57%
• Number of Executed Trades Under/Equal Median Price = 52.34%

For Sells:

• Average Deviation from Median Price = -7.49%
• Number of Executed Trades Above/Equal Median Price = 48.37%

Part of the motivation behind this analysis was to determine if I was adding any value, by performing DOM checks + price selection trades (instead of "Market-Orders") - Trade Executions have consumed a significant summated amount of my time (I've always "tried" to ensure that I get the best value fill).

From my analysis I can conclude that: I have not been adding value and my Trade Price Execution distribution appears to be random.

Any fellow real traders, care to share an analysis on their Trade Price Executions?

I would also like to kick off a discussion on tips/techniques that may provide guidance to obtaining the most efficient Trade Price Executions.

 

[skyQuake]

Why median price vs say VWAP which should be a better benchmark?
I get post trade execution reports from my broker every now and then which shows some interesting stuff like:
What happens after you play a VWAP order? DMA/Iceberg? Aggressive Liquidity Seeking order?
on the 5min? 1hr? 1Day? 30day? timeframes

If analyzing by hand (or using iress fills) I'd highly recommend checking the stats of the orders where you cross the spread vs the ones where you dont.

 

[Trembling Hand]

Yeah you would think that VWAP is the metric you should be checking against. A few ticks at the extremes means little to what you actually can execute at. Not sure where you are going to get that data from historically though? Not without intraday data and your own programming??

 

[skc]

I can see what you are trying to do but I am not sure you are going about this the right way.

The question is: if you don't spend much time with watching the depth and placing orders etc, what would you be doing to get a fill? Your answer seems to be "market order". So you should analyse your realised execution vs market order (at the time which you make the trade). You should not compare your executed price vs median price. It's not something that is meaningful to you.

For a small trader, I think vwap is not that relevant as a benchmark because 1). It's unlikely you need to dice up your order so as not to move the underlying unduly, 2). You probably don't have access to the tool you need to achieve vwap, and 3). Your trading strategy is probably based on certain price level (e.g. support / resistance) that vwap simply doesn't come into it.

Say for example, a simple breakout strategy may trigger a buy when price is >$5.25 resistance. The price opened at $5.15, went thru $5.25 during the day, hit a high of $5.45, closed at $5.20, with a median price of $5.33 and vwap of $5.19. What does any of these numbers matter to you given that you are simply trading the breakout of $5.25? You shouldn't aim for median (as it's a worse price) and you shouldn't aim for vwap (as it is below your breakout level - so unless you pre-empt the breakout or didn't take the initial break and assumed a pull back - you wouldn't have traded that price anyway).

That's not to say you shouldn't look at the depth. In the above example, you could certainly attempt to buy @ $5.24 and avoid $5.26 and $5.27, but that's probably as much as you'd try. In more expensive / more liquid stocks the price of crossing the spread gets smaller, while you should certainly spend the time on a 10c stock and avoid paying 10.5c when you can.

 

[Trendnomics]

Thanks for all the responses guys. Unfortunately, VWAP data is not available.

Just to clarify:

• The whole point of the exercise was to determine if I was adding any value, by placing orders which were based on DOM analysis;
  
• Value adding being: Executed prices are below the Median Price for buys and above the Median Price for sells;
• I'm assuming that the Median Price is correlated with the strongest executable volume;
• I'm assuming that over a statistically significant amount of random trade executions, that the average executed price deviation from the Median Price, would be equal to zero;
• Hence, my current performance almost reflects randomness;
• "Market Order" executions (random) might have delivered the same result with less screen-time (work);
• I'm not a day-trader and I clearly have no predictive capabilities on daily price action - so why am I bothering with DOM analysis?

 

[Trembling Hand]

I'm not a day-trader and I clearly have no predictive capabilities on daily price action - so why am I bothering with DOM analysis?

Yeah but you don't know if it would be worse with a market order. You're trying to do a detailed analysis with a very restricted data set. But you are probably right by guessing that you are adding little to the execution by staring at the DOM all day.

That of course is not to say that you cannot add a lot to your entries and exits. I have no idea how long your trades go for but I wonder if you could get intraday data and run a backtest on a simple oversold signal intraday (below bottom Bol band on 1-5 min or <30 10 Bar RSI etc) for your EOD buy signals. if no execution before 3 pm just go market. Would require some coding though..... ?

 

[Wysiwyg]

• Obtained the High and Low price for the equity on the day of trade execution - I then calculated the Median price accordingly;

I would also like to kick off a discussion on tips/techniques that may provide guidance to obtaining the most efficient Trade Price Executions.

The median price of a daily bar would be the middle and completely random. I mostly use market depth for stock trades on the right and wrong assumption that sizeable volume at a certain level will hold.

 

[Trendnomics]

The median price of a daily bar would be the middle and completely random. I mostly use market depth for stock trades on the right and wrong assumption that sizeable volume at a certain level will hold.

Hmmmm..... appears that some of you are completely missing the point..

 

[Trembling Hand]

@Trendnomics what is the timeframe of your system?

 

[Wysiwyg]

I then calculated the Median price accordingly

Your whole analysis is based on the "median" price and you only stated the high and low on day of execution to extract a median price. Not missing anything.

for example (Executed buy price = 4.51), (Median price = 4.675), (Low price = 4.48) -> (Deviation from Median Price = +84.62%)

+ 84%??? a buy at $4.51 versus $4.675 means you bought 3.66% better off than the random middle number. Again not missing the point.

 

[Trendnomics]

@Trendnomics what is the timeframe of your system?

EOD and I consider myself a bit of a market shifter.

 

[Trendnomics]

Your whole analysis is based on the "median" price and you only stated the high and low on day of execution to extract a median price. Not missing anything.

+ 84%??? a buy at $4.51 versus $4.675 means you bought 3.66% better off than the random middle number. Again not missing the point.

Why are you questioning the 84%? As I've outlined before, this percentage is range bound by the high/low- hence, in this case 0% would be the median and +100% would be the daily low.

Using Median Price for this type of analysis is the simplest approach - way too many hairsplitters on these forums. Just to state the obvious: Median Price cuts the daily range into two parts (upper and lower), if you are buying you would hope for an execution in the lower part, if you are selling you would hope for an execution in the upper part. As my stats show, I'm extremely poor at consistently executing trades in the upper/lower part (as desired).

 

[Wysiwyg]

Why are you questioning the 84%? As I've outlined before, this percentage is range bound by the high/low- hence, in this case 0% would be the median and +100% would be the daily low.

You explained yourself better there and I understand what you mean.

if you are buying you would hope for an execution in the lower part, if you are selling you would hope for an execution in the upper part. As my stats show, I'm extremely poor at consistently executing trades in the upper/lower part (as desired).

If you knew what the upper and lower half of a daily bar were in advance then you certainly would be onto something. Have you considered volume at price where there could be a level that price trades to? For a day or few hold, obviously the entry and exit prices will be preferred at best possible prices. Apart from market depth, support/resistance/VAP I don't know of anything that will assist.
For longer term holds the saying is "Don't be a dick for a tick". Good luck with it all.

 

[Trendnomics]

For longer term holds the saying is "Don't be a dick for a tick". Good luck with it all.

Precisely!!!! This was the point of this analysis. I am a long term EOD trend-follower. But I use DOM analysis to "try" and obtain the best execution prices. I've always maintained the assumption that I was adding value by performing DOM analysis and using screen-time to obtain the best possible execution.

BUT, as the stats are indicating (for over 500+ trades), DOM analysis may have been a complete waste of time - approximated time wasted = 10 minutes x 500 trades = 83.33 hours (that's a fair few $$$ on my current hourly rate).

Perhaps market-order executions (no screen time), could've delivered the same type of statistical outcome (for 500+ trades)?

Anyone else care to perform a similar analysis of their trade histories???

 

[Trembling Hand]

way too many hairsplitters on these forums.

LOL. You come here with a partially/incomplete analysis to ask others opinion then blow off at people looking at the details. You kidding mate?

 

[skc]

EOD and I consider myself a bit of a market shifter.

Given that you trade EOD and don't want to watch the DOM... does that mean you'd just trade the next open? That's my understanding of how an EOD trader typically operates.

If that's true... when you say market order you really meant market-on-open? Or do you do something totally different to what is described here?

I guess I still go back to the point that... to analyse whether your screen time adds any value, you need to compare the price achieved with the correct alternate price. Median price doesn't sound like the alternative that you should benchmark against.

Say your EOD trigger was $1.50, and the stock opened @ $1.50. You have executed perfectly to plan. Suppose the stock ran all day and finished at a high of $1.70 and a median price of $1.60. Your analysis would show that a fill of $1.55 is a value-add because it's below the $1.60 median price - when in fact you are worse off.

P.S. With all the time spent watching the DOM there may be some tacit learning going on... it could be helpful in the future when you change trading style. So hopefully it's not a complete waste of time even if you don't achieve any meaningful improvement in fill.