Testing

This page is designed for people who have acquired a roulette computer and want to test it. Perhaps you purchased one of my computers, or a device from another vendor. By no means are the suggested tests exhaustive, but they cover the main points so that if you were scammed, you will know. If you are doing this testing with my computers, contact me or see the instructions for advice on the correct settings to use.

Same spin testing:

The most basic test involves predicting the same spin on a DVD. As the computer is predicting the same spin, you can expect the predictions to be much the same. This test determines how well the computer deals with errors from manual clicks of the button – nothing more. It is NOT an indicator of a roulette computer’s overall accuracy.

Guidelines:

* Avoid DVD players in PCs as they are notorious for skipping which causes inaccurate timings, and leads to the illusion of poor computer performance. Either use a good quality DVD player or media player

* If you are testing a tilted wheel algorithm, you need to set the target ball speed to no faster than 1000ms (1s per ball revolution). If you are testing a level wheel algorithm, your ball target speed should be 1300 or greater.

* Avoid making any of clicks while the ball is exhibiting sudden deceleration. This will adversely influence results.

* If you are comparing two computers, it is best to join the switches from both computers so that each of them receive the same data, and are subject to the same variables

* Select a test spin with a slow rotor (5+ seconds per revolution) to eliminate rotor timing errors as a factor.

* Use a test spin on a wheel without significant dominant diamonds. If you use a wheel with significant dominant diamonds, the test with level wheel algorithms will be roughly 8-12 pockets less accurate than a level wheel. This is because having the wheel on an uneven surface causes the ball to follow an irregular “wavey” deceleration curve, rather than a more gradual curve (this is because the ball follows a minor uphill and downhill path as it moves around). If you only have DVDs of a tilted wheel (with clear dominant diamonds), then expect results to be significantly worse. But you can still do comparisons between different computers by testing each of them on the same spin, with the prediction at the same ball speed.

Testing tilted wheel algorithms

Tilted wheel algorithms are essentially the basic computer algorithms. They determine nothing more than rotor position at a fixed but approximate ball fall time. There are two main variations to this kind of algorithm. One will include the remaining time in the target revolution and include it with all remaining revolutions, and the other will only consider complete remaining revolutions. You can switch between either options with my Uber computer with the “full rev” setting. I suggest setting it to OFF.

Simply repeat the prediction of the same spin, and note the predicted number. If you use the same reference diamond for timings as you did for the ball sample, you can expect my computer’s basic settings to give prediction roughly 50% of the time. This almost never occurs in real play though because the actual data from play is significantly different. If you want to receive predictions with the same diamond regardless, then use the
“anytime” ball algorithm (seek assistance if you are unfamiliar with it). Vary the times at which you start to clock the ball and/or rotor. Note all the predictions you receive.

After testing about 10 spins, repeat the process but with clocking the rotor on the opposite side for 10 more spins. Clock the ball at the same diamond as you did previously.

Results to expect:

Most predictions should for each set of 10 numbers should clearly be around the same 3 numbers. If this is not the case, re-check the guidelines to see if something is missing. Each different set of 10 numbers should be at opposing sides.

Testing with level wheel algorithms

For this test, set the computer to give predictions for wheels that don’t have significant dominant diamonds. You realistically need 4+ ball clicks to achieve significant accuracy. The faster the ball when you get prediction, the more ball clicks you will need to use.

Now predict the same spin, but each time use a different reference diamond for the ball and green zero. You must use the same diamond for both ball and zero for each trial (each time you replay the spin). With my uber you can use various different settings and see how it affects accuracy. You should have about 5 predictions for each diamond.

Results to expect:

The below data rates the device, and assumes the spin used is on a level wheel, the ball speed is 1300ms per revolution or slower, and the difference in time between revolutions at the target ball speed is 200ms or greater (the larger the difference, the easier it is for the roulette computer to predict):

Excellent: All predictions within same 1-6 pockets

Average: All predictions within same 7-12 pockets

Poor: All predictions within same 13 – 18 pockets

Very poor: All predictions within same 19 – 37 pockets

NOTE: Most of the DVDs I supply have a reasonable tilt, and ball speed difference between revolutions of 90-120ms so they are significantly more difficult to test with. The best DVD to use for this is the V1 DVD.

Comparison of results between Uber and FFA

The only way to properly compare how two computers perform in the “same spin testing” is to join two computers to the same switch arrangement, so they both receive the same timings in the same conditions. The only roulette computers able to achieve reasonable results for level wheel testing (same spin) are Forester’s FFA, my Uber version, and my Hybrid device for automated rotor and ball tracking. The hybrid uses automated image recognition equipment, and predicts the same number almost every time so it isn’t a fair comparison. For a comparison between the Uber and FFA computers with level wheel same spin testing, see the below video:

Your browser does not support this video format

See a full analysis of the results of this test

Ball scatter on different rotor speeds

For this test you will need your own roulette wheel. Tilt your wheel so it has a strong dominant diamond. You do not want the diamond to solidly deflect the ball down – the ball should have an average diamond hit where the ball hits the rotor at an angle of about 45 degrees. If you find this hard to achieve, you will have an older and/or easily beaten wheel that is unsuitable for the test. Test your computer over 100 spins with the rotor speed at about 5 seconds per revolution. Note the scatter charts. Then predicting at the same ball speed (just use same ball sample data), test over another 100 spins but with a rotor speed of about 3 seconds per revolution. You will notice the scatter charts are significantly different to the extent that you cannot use one of the scatter charts to predict where the ball will bounce. On most wheels, a variation of even just 1000ms (difference between 4 and 5 seconds per rotor revolution) is enough to make the ball travel over 18 pockets the difference, and often give a completely different peak formation. This shows the accuracy you lose with the basic roulette computer algorithm, and this is not even considering ball deceleration rate changes, so you can better understand how unsuitable basic roulette computers are in real casino conditions.

I suggest you record the spins so you can later use the Uber or Hybrid’s features to adjust for the variation in scatter on different rotor speeds.

Especially if you use a basic algorithm, you need to carefully observe the number of remaining revolutions after each prediction to ensure they are consistent, and haven’t been varied from ball deceleration rate changes. If you very heavily tilt your wheel and obtain very late predictions (about 3 seconds before the ball actually falls), deceleration rate changes will be far less significant. This enables you to focus only on rotor and ball interaction. Also be aware even with the same number of remaining ball revolutions, the change in ball deceleration does affect the time at which the ball hits the dominant diamond. For this reason, also observe the unadjusted predictions when the ball meets the dominant diamond.

Ball deceleration rate changes

For this test you will need your own wheel. Set your computer to give predictions about 10 seconds before the ball falls. Have the rotor speed at about 4s/revolution. If the ball is way too fast when released, and suddenly decelerates, your ball track would be very worn, and the wheel is too different to real casino wheels to do this test. Calibrate the wheel so that you have a reasonably clear single dominant diamond.

Purchase any reasonable barometer watch that measures air pressure. Set the default height to 0m (or 0 feet) altitude. Test your roulette computer over 100 spins and note the location of peaks on your scatter chart. Now wait a day or so, or even later in the day for the altitude to be different by about 20m. Now re-test over 100 spins, with exactly the same ball sample data and settings used. You will notice the peaks from your scatter chart have significantly moved. The movement in the stated conditions is approximately 9 pockets. If wheels with weaker dominant diamonds, the effect is much greater.

This all occurs because of the air pressure change that constantly occurs, you have gone from targeting the correct area of the wheel to avoiding it.

Further testing:

Try the test on a wheel with both one then two or more dominant diamonds. You will find the ball deceleration rate change can transform one clear peak into two smaller peaks, or the opposite. Often you may find there are three peaks that are significantly smaller and less clearer than previous results.

There are far more factors that affect ball deceleration rates, but the point of this test is to see how much a simple air pressure change can affect a basic roulette computer algorithm (see mathematics) that doesn’t adjust to ball deceleration rate changes. My Uber and Hybrid roulette computers are the only roulette computers capable of adjusting to deceleration rate variations. The adjustments can be done either automatically or manually.