93% Prediction Accuracy – How Can I Profit From This?

[davemake]

Hi Folks, as described in the title, I am trying a method of forecasting future High/Low levels of various instruments.

In essence, the system looks at EOD High, Low and Close data and then provides a forecast that the Index/Share price will break through one or both of the forecast Upper or Lower price levels over a specified period (4, 9 or 15 days).

I then trade these levels (see below)

Accuracy (Measured over 100 trading days) is quite good – best 93% (CAC40), worst 76% (DJ Index) over the 4 Trading day forecast.

The forecasts are ‘time period’ specific, (4 days, 9 Days and 15 days).

--------------------------------
Example 1
Date: 01/12/08
Index = CAC40
Open = 3261
High = 3272
Low = 3072
Close = 3080

4 day forecast – derived from the above data:-
CAC40 will hit either 3372 or 2972 on or before 05/12/08 (4 day forecast)

# Result: CAC40 broke through the 2972 level on 05/12/08 – Day 4 (CAC40 low was 2957)

3372 was never reached over the 4 day period up to 05/12/08
--------------------------------
Example 2
Date: 06/01/09
Index = CAC40
Open = 3373
High = 3426
Low = 3339
Close = 3396

9 day forecast – derived from the above data:-
CAC40 will hit either 3513 or 3253 on or before 19/01/08 (9 day forecast)

# Result: CAC40 broke through the 3253 level on 12/01/08 – which was, in fact, Day 4 (CAC40 low was 3236), and continued downwards over the next few days of the time period.
(In this case, the lower level was hit well within the 9 day specified period.)

3372 was never reached over the 4 day period up to 05/12/08
--------------------------------
Example 3
Date: 15/01/09
Stock = RDSA – Royal Dutch Shell (Accuracy = 83% on 4 day forecast)
Open = 1692
High = 1718
Low = 1658
Close = 1669

4 day forecast – derived from the above data:-
RDSA will hit either 1748 or 1628 on or before 19/01/08 (4 day forecast)

# Result: RDSA broke through the 1748 level on 19/01/08 – Day 2 (RDA0 high was 1763.

1628 was never reached over the 4 day period up to 19/01/08.


I have more examples if required. On many occasions, both Upper and Lower limits are broken over the fixed time period

My problem is that the only means of trading these forecasts are using the BetOnMarkets - The smarter way to trade web site. I use the “Up or Down” bet on the “Boundary Bets” section.

The other site I use is http://www.betsfortraders.com (One Touch Range Bet).

I’m looking to increase my returns by using Spread Betting as an alternative to the above web sites.

I'm seeking the best strategy in order to acheive an increase in returns.

All suggestions will be appreciated, Thanks in advance !!

 

[pedro01]

Dave

Depends on the amount of $$$ you have to invest & how many times you will trade per week.

With the DJIA - you can open a futures account to trade on the index - look at about $5K to have on balance with the broker as a minimum. For the individual equities, if you trade the US markets and you are going to have in excess of 4 round trip stock trades in any 5 day period you'll need to maintain a minimum balance of $25K in your account. You don't need this much if you trade less frequent.

If your system is purely technical, then you can scan a large list of stocks for signals in Tradestation and have it place the trades on qualifying stock for you automatically.

One thing though - 100 days is not enough for backtesting, and 76% win ratio means nothing on it's own - there's other metrics that you need to get on the results that tell you if it's profitable.

I'd do this:
1 - get Tradestation or some other trading software
2 - get your system into Tradestation
3 - backtest it on years of data across various indices & stocks
4 - get the results metrics to see if this system is truly profitable
5 - if it is - go live

This is something I can help you out with as long as the system isn't overly complex. I'm pretty busy working on another system right now.

The market has been through a pretty volatile time over the past 100 days and I have a breakout system that shows huge gains, especially in that time period. Prior to that it is still profitable in the previous 2 years but nothing like the gains in the past 100 days.

Cheers

Pete

 

[Dacamic]

Pete has offered very good advice. To re-emphasize one of his points, you should definitely backtest your system across more than 100 days and in a variety of market conditions.

 

[davemake]

Pete, Dacamic,

Many Thanks for your very useful advice, I will perform further back testing and investigate/supply the results.

Because there are 2 possible outcomes of the forecast how could I trade futures? Would not a 'Buy' trade profit cancel out a Sell trade profit? I.e. 1 point gain on a Buy contract would reflect on a 1 point loss on the corresponding Sell contract.

Referring to my first post:

My problem is that the only means of trading these forecasts are using the BetOnMarkets - The smarter way to trade web site. I use the “Up or Down” bet on the “Boundary Bets” section.

The other site I use is http://www.betsfortraders.com (One Touch Range Bet).

I’m looking to increase my returns by using Spread Betting as an alternative to the above web sites.

I'm seeking the best strategy in order to achieve an increase in returns.

 

[pedro01]

Pete, Dacamic,

Many Thanks for your very useful advice, I will perform further back testing and investigate/supply the results.

Because there are 2 possible outcomes of the forecast how could I trade futures? Would not a 'Buy' trade profit cancel out a Sell trade profit? I.e. 1 point gain on a Buy contract would reflect on a 1 point loss on the corresponding Sell contract.

Referring to my first post:

My problem is that the only means of trading these forecasts are using the BetOnMarkets - The smarter way to trade web site. I use the “Up or Down” bet on the “Boundary Bets” section.

The other site I use is http://www.betsfortraders.com (One Touch Range Bet).

I’m looking to increase my returns by using Spread Betting as an alternative to the above web sites.

I'm seeking the best strategy in order to achieve an increase in returns.

Don't worry too much about the specific order types. You can use an OCO order - linked orders where if one gets hit, it cancels the others.

So - you could have 2 limit orders, OCO and when 1 limit order opens, the other one gets cancelled. Limit orders won't be placed until the price you want to get in at gets hit.

I'd worry more about the backtesting right now.

 

[horace]

Hi!

Further on Pedro's advice (which I fully susbcribe to) I'd anticipate a couple of things.

Be advised that the winning % is the least important of a tradings system metrics. You can manage it as pleased as long as you know how. Frequently a high percentage winning system is attractive. What people don't realize is that a high percentage winning rate means (almost always) extremely high risk of ruin.

Also to note that if you are going to trade this system in the open markets you NEED to place a stop loss in your trades. Please DO NOT trade this (or any) system without a stop loss in place. If you do, sooner or later you're going yo realize it was a bad decision. I say this cause placing a stop loss almost gaurantees your % percentage drops dramatically (which is good).

A rule of thumb for the winning percentage is (al least for me) never trading a system with more than 60% accuracy. Frequently under 55% is enough. And I live on revenues of systems under 40%.

If you need any help dont hessitate...

Horace