Random Trader

[Effkay]

So here I am a bit obsessed with random trading at the moment, but here goes!!!

I got a spreadsheet of historical S&P eod data going back to novemeber 2002. I put a random 1,0 next the data (using rand()). This would make the trader a buyer or a seller at the close. All positions were then closed at the following day's close, and a new position taken. Here are the results of 3 traders following such a strategy. Starting with a £10000, and betting £10 per point, and allowing for £10 lost in the spread

Trader A: £5361.70
Trader B: £7088.24
Trader C: £4274.46

What's my point?

My point is that once I had accounted for spread, I had not one single profitable trader over the period of 2 and a bit years. (Trust me I pressed F9 pleanty of times!!)

So taking this into account. there was no way a random trader could have made money in the long run!!! (A surprise result for me!!!)

 

[trendie]

this is a profound disappointment to me.

the very foundation of my beliefs have been undermined.

oh well, maybe God doesnt play dice...., even with the stock market.

have you tried runes ?

 

[Effkay]

I always thought that it was possible for someone to be very very lucky and make money in the markets. That's all!

Having said that if I apply money management to the very same results I wonder what would happen? Any ideas?

 

[Tuffty]

If there were say approx 500 trades, that would have cost 500x£10, so two of the traders would have made money if there were no costs in trading. The result of random trading should be a normal distribution curve with some very lucky big winners and some not so lucky loosers with the majority falling around break even (excluding costs) which does appear to have happenned in your small sample of three.

 

[gcb01]

Something Similar

I did something similar with an FTSE model but betting both ways each day. The outcome is that with stops set fairly tight then if volatility is high enough ( it hasn't been for a while) then the model makes money.

 

[Effkay]

gcb01: how does it make money if it bets both ways?

 

[gcb01]

Easy

gcb01: how does it make money if it bets both ways?

It the market is volatile and you arer getting say 100 points up or down each day and you set stops at say 30 points then you get +100-30-spread-commission. Doesn't happen every day and depends on high volatility. I modelled it on EOD. As the model showed that volatility levels had fallen to the extent that this did not work, I never attempted to try it for real.

The model did assume that you exited at close and did allow for both long and short stops being triggered (a bad day).

One of the keys to making it work was adjusting stop levels in line with volatility.

What put me off it was that apart from not making money (currently) due to lower volatility, it makes loads of dosh for your broker.

 

[Effkay]

Surely you could trade that differently:

Use a breakout, if the market moves up +30, enter long, if it moves down -30, sell short?

 

[gcb01]

Eod

Surely you could trade that differently:

Use a breakout, if the market moves up +30, enter long, if it moves down -30, sell short?

Indeed you could but I only had EOD data to model on.

 

[ardhill]

With the idea from gcb01, how about this?

- Take an instrument of your choosing.
- At the EOD, calculate the difference between the high and low eg. 130 pips/ cents/ pence or whatever
- Lets say that this is now your profit target and your stop loss is half of that.
- Enter an OCO boxed order on the breakout of the high or low eg Buy the next day 1 c above the high, or short 1c below the low.
- Using our example above, your stop would be 75 and your profit target would be 130.
- If a trade is executed, then stay in it until you are taken out with your stop or limit order.
- If a trade is not taken that day, then set your new parameters from that days high & low for tomorrow.

I would try this on a spreadsheet, but I don't have a clue how to go about it, come to think about it, I use TradeStation and I don't even know how to program it in that to back-test the trades

Just a though for anyone interested in trying a slightly different random trading test, that has a better programming ability than me (not that hard).

If you decide to have a go at programming this, then you may even want to play with the stop eg instead of half the high / low spread, maybe 1/3 or 1/4?

 

[Effkay]

yeah that's similar to the stops and limits i've got on the dice trading thread, although I use ATR14 as I don't want wide days and narrow days to mess up the trading.

 

[gcb01]

With the idea from gcb01, how about this?

- Take an instrument of your choosing.
- At the EOD, calculate the difference between the high and low eg. 130 pips/ cents/ pence or whatever
- Lets say that this is now your profit target and your stop loss is half of that.
- Enter an OCO boxed order on the breakout of the high or low eg Buy the next day 1 c above the high, or short 1c below the low.
- Using our example above, your stop would be 75 and your profit target would be 130.
- If a trade is executed, then stay in it until you are taken out with your stop or limit order.
- If a trade is not taken that day, then set your new parameters from that days high & low for tomorrow.

I would try this on a spreadsheet, but I don't have a clue how to go about it, come to think about it, I use TradeStation and I don't even know how to program it in that to back-test the trades

Just a though for anyone interested in trying a slightly different random trading test, that has a better programming ability than me (not that hard).

If you decide to have a go at programming this, then you may even want to play with the stop eg instead of half the high / low spread, maybe 1/3 or 1/4?

It's not that difficult to set up but with EOD data you have to make some (heroic) assumptions such as:
1) what price to you get on exit, close or high/low, depending on which bet triggered and
2) what you do if both bets are triggered, i.e. which went first.

I'll have a look at it .

 

[Edster]

Trader A: £5361.70
Trader B: £7088.24
Trader C: £4274.46

What about a wider sample of traders - say 10?
Also, what if your hypothetical random traders set stops of say 10 or 20 pips for each trade rather than being down up to 100 pips a day?
That'd be interesting to see what happened to your final figures.

Edster

 

[henry766]

not really

 

[FetteredChinos]

yes way

 

[ardhill]

It's not that difficult to set up but with EOD data you have to make some (heroic) assumptions such as:
1) what price to you get on exit, close or high/low, depending on which bet triggered and
2) what you do if both bets are triggered, i.e. which went first.

Point 1 - Are you asking what price you exit at when the trade is triggered? If so, maybe I didn't explain myself properly, sorry.

Let me give you an example. The price range today for Amazon (AMZN) trading on the Nasdaq was High - 42.16, Low - 41.01. The price range was therefore, High-Low= 1.15

The trade tomorrow would be, if AMZN went to 42.17, then we would go long.
Our profit target would be 43.32, which is 1.15 from the position (the previous day's range)
Our stop would be at 1/2 of the days range away, lets say 57 cents so 42.17 - .57 = 46.60
We would stay in the trade, overnight if needs be, until the trade was taken out by one of our orders - stop loss or profit target.

If the price tomorrow on the other hand went below todays low @ 41.00, then we would go short.
Our profit target would be 1.15 away, and our stop loss would be 57 away.

Only one trade can be triggered in a day, and the trade says in the market until it is taken out by one of our orders. So we will be up by 1.15 or down by 57.

Lets say tomorrow a trade is taken and is closed on the same day, then tomorrows range would be the calculation. High-Low = target, and half the target is the stop loss. The new orders would be entered, 1 cent below the new low for a short, and 1 cent above the new high for a long position.

Does that make sense?

Point 2 - Only one trade is triggered a day, I one is taken, the other order is cancelled - OCO

 

[gcb01]

Results

ardhill,

Results below. Doesn't look to good. Any suggestions?

Currency: Points
Long
Gross -215,932.06
Commission/Slippage -6,560.00
Net -222,492.06
pts per day -45.73

Trades 1312
pts per trade -169.58

Short
Gross -117,215.08
Commission/Slippage -5,855.00
Net -123,070.08
pts per day -25.30

Trades 1171
pts per trade -105.10

Total
Gross -333,147.14
Commission/Slippage -12,415.00
Net -345,562.14
pts per day -71.03

Trades 2,483
pts per trade -139.17

First Day 30/09/1985
Last Day 10/01/2005

Commision/Slippage 5 points per trade

 

[ardhill]

ardhill,

Results below. Doesn't look to good. Any suggestions?

Net -222,492.06

Yes, I believe my local council are looking for new Bin Men, I think I might apply - time for a new profession

Well that slaps the idea of going with a breach of the previous day's high or low... Or does it?

I must admit, non-discretionary management and trading limits are an area which I have never had much success. I can find mechanical ways of getting into a trade, but as for getting out, I have had to use err, personal trading skill.

I still reckon the idea of getting into the trade is good, but management is something else.

Where are the people on T2W who are good at this mechanical system stuff - come on you know who you are, how can we manage this so that it is profitable?

As for purely random trading

Looks like some skill might still be involved

 

[rog1111]

1 pt spread seems rather high ? 0.25pts is the norm for ES.

rog1111

So here I am a bit obsessed with random trading at the moment, but here goes!!!

I got a spreadsheet of historical S&P eod data going back to novemeber 2002. I put a random 1,0 next the data (using rand()). This would make the trader a buyer or a seller at the close. All positions were then closed at the following day's close, and a new position taken. Here are the results of 3 traders following such a strategy. Starting with a £10000, and betting £10 per point, and allowing for £10 lost in the spread

Trader A: £5361.70
Trader B: £7088.24
Trader C: £4274.46

What's my point?

My point is that once I had accounted for spread, I had not one single profitable trader over the period of 2 and a bit years. (Trust me I pressed F9 pleanty of times!!)

So taking this into account. there was no way a random trader could have made money in the long run!!! (A surprise result for me!!!)

 

[Rhody Trader]

I actually ran a test on the S&P a while back similar to the ideas presented in this thread. The system was to enter an options straddle (simultaneous put and call in an at the market strike) with a fixed profit target on each side, like 10 points. Hold for a week. The system would have done very nicely up through about 2002. Since then however, not so good. The drop in volatility the last couple of years (using ATR as the measure) has negatively impacted the systems profitability. Maybe it will come back when the ranges start getting wider.