Most of the time a longer optimization period produces better results. 500 to 1500 days often work well. Unfortunately, all stocks do not have such a long period of historical prices available and you are forced to use shorter periods. Other alternatives are to set the parameters manually or to use Optimal Trader's default parameters. Be careful not to overfit the models. Optimization results often give a better impression with shorter periods compared with longer periods, but longer periods produce more reliable signals because the models will be trained with wider range of scenarios.
If a stock has an entirely different price development pattern a long time ago compared with today it may be better to use shorter optimization periods.
Read more in the section Technical analysis with optimized parameters.
If you set the brokerage fee to 1%, then 1% is charged at every trading occasion.
Some funds do not charge an initial fee when buying fund shares, but only charge their shareholders when they sell shares (redemption fee). If that is the case then set the brokerage fee to half the redemption fee, because the redemption fee is only drawn at every second trading occasion.
So if your redemption fee is 1%, then set the brokerage fee to 0.5%.
The optimization period defines how long time backward is used when optimizing the parameters of the models. If the optimization period is 250 days it means that the models are tested the last 250 days in order to find the best possible parameter combination. This optimal setting is used to produce a trading signal(buy or sell) for today.
Back Testing simulates use of Optimal Trader through carrying out this optimization day for day a number of days backward in time to get daily trading signals. Back Testing thus provides you with an answer of how well the models have worked in reality.
The optimization period that is used each day in Back Testing is the same as the one used for normal optimization.
The numbers days you set in Back Testing define the number of days the simulation is carried out.
Optimal Trader mostly provides historical prices adjusted to splits and dividends. To get the adjusted prices you have to redownload the specific stock because the prices are not automatically adjusted on the client computers. First remove the stock from the portfolio, and then download it again to get the adjusted prices from our server.
If historical prices are not automatically adjusted you can adjust them manually using the Options/Split Adjustment function or editing the price data in Excel.
This is easiest explained with an example. Take a look at the chart below.

Take a look at the chart of the previous question. If the date would have been displayed at the last price value, the graph would have been forced to be a bit smaller. We have given priority to as large charts as possible and thus we have chosen not to show the last date in the charts. There are two easy ways to see the date for the last price value:
You must update the lists in the Add dialogue to obtain the full lists. Click on Update under a Sub Category and you will have access to all equities in that market. This must be repeated for every market you are interested in.
Use the Portfolio Scan feature of Optimal Trader to find winning equities based on your investment criteria and neural network forecasting.
Read magazines with stock analyses to find interesting investment choices and examine how well the stocks are suited for trading in Optimal Trader.
You can also use fundamental analysis to find stocks based on your investment criteria. There are several free online functions to filter and sort stocks based on fundamental facts.
When performing standard optimization, parameters in the system are tested with different values and the values that have produced the highest profit are the ones used. If you use long optimization periods standard optimization often produce stable and profitable systems, but there are more reliable ways to optimize the parameters.
In Optimal Trader you have the possibility of optimizing the models so that they are trying to produce high and stable profit instead of trying to maximize the total profit. The result is that the probability of profitable future results increases. This option is called Robust Optimization.
If you do not want to use optimized parameters you can use Optimal Trader's default parameters which are selected to generally to produce reliable and stable results, or set the parameters manually.
Read more in the section Technical analysis with optimized parameters.
Neural networks are not deterministic systems, but are initiated from a random initial state and are trained from that state to make conclusions about the training sequence( price history).
It is possible to make a deterministic system out of a neural network, but it is not advisable. This is because the training sometimes fails and you would always get the same results when training the network again if it was initiated from the same initial state.
If you select Model Settings/ANN and check Deterministic Training, the network will always be trained from the same initial state and you will always get the same results.
It is possible to obtain more consistent results at a cost of longer processing time by adjusting the following parameters:
1. Increase the Number of Networks - Because every single neural network produces slightly different results, the results of several networks are combined to produce a more stable result. The value of this parameter sets the number of networks used for the combined result.
2. Increase the Number of Epochs - Determines how much the network will be trained. Please observe that high values may result in curve fitting.
This discussion only applies to neural networks. Other models in Optimal Trader are deterministic systems.
The neural network uses both price history and technical indicators as inputs, but does not use the results of the active models. It is better to let the neural network find patterns by itself than feeding it with data that is already adapted to price behavior. Such a net would not become stable and general. It could find a “trivial” solution such as to entirely follow one of the active models.
The models and indicators of Optimal Trader always generate either a buy or a sell signal. The coloured fields and the signal of today are either green or red, so any uncertainty can not be detected in the price charts or the trading signals of today. If there is an uncertainty, the signals can change often, historically and even intraday. Only looking at the coloured signals is not enough in these situations.
In the enlarged layout you can evaluate the strength of the trading signal. If the signal for a model is close to a limit, the trading signals can vary a lot. This should be seen as a sign that the signal is not strong and the model is having difficulty drawing conclusions.
Example: If you double-click the ANN-chart you can see in the lower chart that the model curve is oscillating between 0 and 100. The limit between a buy and a sell signal is 50. If the model curve for the neural network is moving between 45 and 55 for a period of time, the trading signals may change often, but in reality the difference is very small. Thus, values close to the limit indicate uncertainty. If you never take a look at the model curves you will not see this uncertainty!
Every model is different and different model curves must be understood in different ways. Read more about the different models in the section Technical Analysis Models and Indicators.
See the answer to the previous question.
If you optimize the parameters of a model a specific day, parameters producing best results the last 1000 days are generated (if you haven't changed any settings). It is also for these parameters Optimal Trader is showing historical trading signals by means of the coloured fields. These parameters are usually quite stable if you use long optimization periods. But if you optimize the same model a day later it may sometimes happen that Optimal Trader finds another parameter combination which has produced better results because the basis for the optimization has changed. Historical trading signals will also be changed as a result of changed parameters.
Follow the instructions below to backup your data and use it on a new computer.
On your old computer:
1. Start Optimal Trader, click Options and select Make Backup.
2. Save your backup file (*.backup) to a USB-memory stick.
On your new computer:
1. Install Optimal Trader on the new computer and start the program
2. Click Options and select Restore From Backup.
3. Follow the instructions and select the backup file from the USB memory stick
4. Select Options/User Key to create a new User Key and send it to us.
Wait until we have activated the new User Key after which you can use Optimal Trader again.
You can zoom into a region of a chart by clicking and dragging with the left mouse button in the chart. The time periods of the other charts are automatically updated to the same region.
You can check the version of your copy under Options / About Optimal Trader.
Yes, the trial version is exactly the same software as the full version. In fact when you purchase Optimal Trader, we will unlock the trial/evaluation version to a full version. No re-installation is required.
You should re-optimize the models regularly to keep the models adapted to the current price behaviour. You should also run Best Models and Settings regularly for the same reason. Best Models and Settings results are based on daily re-optimizations.
You can approximate this by calculating a rough percentual fee. For instance, if the price for a share is around $4.2 and the fee $0.01/share, the percentual fee would be 0.01/4.2 = 0.24%. This approximation will not have a noticeable influence if the price is not changing dramatically over the optimization period.
Yes. It actually works this way. But when you are using an optimization period of x days and you only have x number of days of data you can not backtest the data a single day. There is not enough data to do that. The only thing you can do is to reduce the optimization period and try again or wait until you have more data of the equity allowing you to backtest the data more.
We have integrated our risk management tool and our scanning tool. You can let the expected returns in the Portfolio Optimization window be proportional to Portfolio Scan results. The result of Portfolio Scan is a value for each stock which will be larger the better the stock has performed with regard to your criteria. Portfolio Optimization will allocate the assets in your portfolio with regard to Portfolio Scan scores of each stock, the risk of each stock and correlation between stocks. Besides that you can automatically deselect stocks from the portfolio if they currently do not have a buy signal.
Scanning thousands of stocks with regard to some criteria for the top 10 stocks often returns strange results. It is statistically better to scan a hundred stocks causing the likelihood for extreme price behaviour to drop radically.
Hurst Exponent estimation needs much data. When you use a small optimization period, maybe 200 days, the estimation will on average deviate more from the Hurst Exponent value, compared to when using a longer timer period to estimate the value.
Yes, it has forecasting value, but somewhat limited. Step-by-step backtesting produces estimations of the expected returns on average, but the variance is large and the results are not conclusive enough. It is better to use a strategy which has been tested on many different equities under different conditions. Use the best active models and the best optimization settings based on empirical results for the current price behaviour pattern. Using the Best Models and Settings feature will allow you to base your investment decisions with the statistical best indicators and settings.
You should only use strategies which statistically have been proven to work for the current price behaviour. Sometimes it is the ANN, sometimes other models work better with specific settings. Using the Best Models and Settings feature will allow you to base your investment decisions with the statistical best indicators and settings. Check the generated report and decide which model you would like to follow.
The Combined Analysis Model signal is the weighted average from the signals of the active models. The weighting constants are determined during optimization, but can be manually adjusted.
Although Optimal Trader provides 15 minutes delayed data, The Optimal Trader system is not optimized for intra-day daytrading. It is designed and optimized for trading in a longer perspective, from some days and upwards.
Yes, under the Time Settings groupbox within the control collection on the right side of the charts you can select an earlier end-date in the To box. The charts will be updated according to your settings. Optimization will be performed on a period up to the selected date. The optimized parameters will be used until you re-optimize the system. After optimizing you can thus switch back to the current time period and analyze the performance of these parameters for the current time period.
Yes, that is correct. You can select to work with absolute returns instead (under Options)
Optimization results are the results of a simple back test, i.e. you look at how a certain model would have performed backward in the time with some parameters. It is easy to find good parameter values for a period afterward, just as it is easy to realize how you should have traded a stock afterward. The goal of optimization is to find optimal parameters, not to give an estimation of future returns.
Return values in the Best Models and Settings report are an unbiased and more realistic indication of a model's strength with the current price behaviour.
Returns generated by optimization are simply the returns obtained when using the set of parameters selected by optimization. Many customers are selecting model parameters manually, testing their own strategies, and want to evaluate the results. Showing the returns obtained when using a certain set of parameters are helping them do evaluate their strategies. The results of manually tested parameters are a better indicator of future returns than parameters generated by optimization, even if that seems strange. This is a statistical phenomena which is thoroughly explained in Curtis Faith book 'The Way of the Turtle'. This book is recommended if you want to get a deeper understanding of optimization and backtesting.
Take a look at price behaviour before and after summer of 20XX. The behaviour before the summer is very different from the behaviour afterwards. Parameters trained to the behaviour before the summer of 20XX are probable not working today when the price exhibits stronger volatility and shorter trends. That is why optimizing more often is a good idea – that way you can take into account the latest price movements.
Take a look at the section Technical Analysis With Optimized Parameters describing similar problems as yours. See especially Warning 4 in that section.
Try setting 5% as a weight limit for a stock. Then right-click that value in the Weight Limit column (5%) and select Apply value to the whole column.
You will see that the result is a portfolio where the smallest weight will always be larger than 5%, but for some stocks the weights will be 0%.
If you would have set 5% as Min Weight, then all stocks would have been included in the portfolio with a minimum of 5% allocated for each stock.
If your trading horizon is one week, then expected returns are not estimated for the next year, but for the next week (If your trading horizon is one month, then expected returns will be for the next month and if it is one day expected returns are estimated one day ahead). Expected return estimation is presented as an annualized return for all trading horizons.
Expected returns/year are thus not returns that may be expected for the next year, but for one trading horizon period from today, but presented in an annualized form.
Expected returns for all different scenarios are presented here. As you can see, Optimal Trader is profitable on average in most market states. In some states it is much more profitable on average than in other states.
If expected returns are low, it is because information in the price cannot currently be used in a profitable way. Next week, price behaviour may be completely different, and indicate more profitable returns.
Also, remember to use the Best Models and Best Settings-feature giving you the highest expected returns.
Yes, you can specify proxy settings, but not in Optimal Trader directly.
A local computer may specify that a default proxy be used. If no default proxy is specified on the computer, proxy settings inherited from Internet Explorer on the local computer will be used. If there are no proxy settings in Internet Explorer, the request is sent directly to the server.
| Copyright © 2005-2010 Optimal Trader |