I've just about recovered from my illness. Fascinating really. More out than ever went in (sorry). Anyway I'm back into things and aim to at at least be up to date on emails by the end of today.
The new roulette computer website is online. If you wanted to have an ad published in the used computer section, let me know.
The online software programmers have finally agreed to fix the simulation bug, but after that is done I'll still change programmers.
I'm way behind on shipping but this will all be sorted this upcoming week.
Regarding the Uber, the programmer is still learning the source code. I will receive an update on his progress shortly. When he's finally past this part, he can finally make the small mod then I can resume shipping Ubers.
And finally the Hybrid, I'm expecting the latest build in the next day or so.
Friday, March 30, 2012
Wednesday, March 28, 2012
Limited responses for few days
Unfortunately I've caught a nasty bug (begins with G and isnt nice) so I'm far from running at my best. I've responded to all critical issues like code requests. I'll still attend the support forum and get some emails done, but I'm quite broken for now so please be patient and I expect to be up to date on all emails by the end of the week. I'm aware of a lot of emails in my inbox, and will get to them asap - you dont need to resend.
Thursday, March 22, 2012
V4 DVD with uber
For players testing the computers on the #4 DVD, that video is a perfect example showing:
- a good set of dominant diamonds, with bad scatter
- How variation of rotor speed affects ball bounce
- How gradual ball deceleration rate changes affect outcomes
For testing, I suggest use the basic auto target, so don't aim for specific diamond hits. Check the raw chart which shows the basic computer algorithm. Keep comparing it to the COM chart for target. With correct application you will find that the raw chart starts out with great accuracy, but after about 40 spins the accuracy turns terrible. But the accuracy is maintained on the COM chart with target. If you do it right, after the 110 or so spins, you will see the difference between almost no accuracy with basic settings, and about 80% wins covering 18 numbers.
Uber Dynamic Tuning & General Update
Many variables determine how the ball's bounce will vary betweren different rotor spins. On a set of spin dvds sent to me by a player, the ball travelled 19 pockets extra when the rotor was just 1 second faster. So if the rotor is just slightly faster, the ball will biounce completely differently. On such a wheel, it would be just about impossible to achieve any accuracy unless you used dynamic tuning, which adjusts for the different ball bounce on different rotor speeds. The lite and basic computers don't do this, same as devices from other developers. Even software used by casinos to assess wheel vunerabilities don't have this feature -= they all assume the ball will bounce the same regardless of rotor speed which is obvious nonsense. It is a feature only user by the uber and hybrid versions.
I also did further testing on the news uber modification, but this time there wasn't a significant improvement (about the same results as target). This is because of a ball and rotor combination anomaly that sometimes occurs. You cant always predict when or it it will occur though without collecting 60+ spins per direction. I'll continue work on it but again it will be months before this is fully implemented.
As for the uber and the new programmer, he's still learning the code but told me he expects to be finish about mid next week. Then he can start coding.
I also did further testing on the news uber modification, but this time there wasn't a significant improvement (about the same results as target). This is because of a ball and rotor combination anomaly that sometimes occurs. You cant always predict when or it it will occur though without collecting 60+ spins per direction. I'll continue work on it but again it will be months before this is fully implemented.
As for the uber and the new programmer, he's still learning the code but told me he expects to be finish about mid next week. Then he can start coding.
Wednesday, March 21, 2012
Calls tomorrow
I'll be taking calls tomorrow but please only call if the matter cannot be resolved via the support forum or email.
Uber and Hybrid development
I've done more research into the possible new algorithm/setting. In the last test, the results were:
Basic algorithm: near no accuracy
Target algorithm (no dynamic tuning): +122 units
New algorithm: +220 units
This was a similarly very difficult wheel. So again it showed a big improvement in the results, but it was more the data the confirmed the improvement.
Testing is one thing, but implementation to make it a practical setting is going to be very difficult. It is a lot more complicated than I expected especially when incorporating the dynamic tuning algorithms (different ball bounce on different rotor speeds).
To do it all in the optimal way will need a 3 dimensional chart using 3 dimensional areas to calculate predictions. A normal pc can do it, but I'm not yet sure if the mobile phones are capable of it. And no off-the-shelf microprocessor can do anything close to it. The phones we use have the storage, but may not have the RAM (memory). So perhaps only the hybrid will be able to do it, unless we can either do further firmware modifications or code the algorithm to use minimal resources on the phone.
Nevertheless, first for the uber we'll get the new programmer working on the firmware compatability issue so I can resume shipping. Within 2-3 weeks I will have designed the algorithm, then we can start coding it. Coding will take roughly a month so it will not be ready for perhaps 2-3 months. Only at this time I will know if it is viable for the uber, or only the hybrid.
The current uber version will do part of it when we re-enable the required menu items, but it is only a limited version of what will be possible. So users can test the improvements, but it is quite fiddly to do compared to what it will be.
Basic algorithm: near no accuracy
Target algorithm (no dynamic tuning): +122 units
New algorithm: +220 units
This was a similarly very difficult wheel. So again it showed a big improvement in the results, but it was more the data the confirmed the improvement.
Testing is one thing, but implementation to make it a practical setting is going to be very difficult. It is a lot more complicated than I expected especially when incorporating the dynamic tuning algorithms (different ball bounce on different rotor speeds).
To do it all in the optimal way will need a 3 dimensional chart using 3 dimensional areas to calculate predictions. A normal pc can do it, but I'm not yet sure if the mobile phones are capable of it. And no off-the-shelf microprocessor can do anything close to it. The phones we use have the storage, but may not have the RAM (memory). So perhaps only the hybrid will be able to do it, unless we can either do further firmware modifications or code the algorithm to use minimal resources on the phone.
Nevertheless, first for the uber we'll get the new programmer working on the firmware compatability issue so I can resume shipping. Within 2-3 weeks I will have designed the algorithm, then we can start coding it. Coding will take roughly a month so it will not be ready for perhaps 2-3 months. Only at this time I will know if it is viable for the uber, or only the hybrid.
The current uber version will do part of it when we re-enable the required menu items, but it is only a limited version of what will be possible. So users can test the improvements, but it is quite fiddly to do compared to what it will be.
Tuesday, March 20, 2012
Software bug
The online software programmers modified something and introduced a bug I only just found. It seems to affect the simulation feature. It doesn't seem to affect betting charts though. You can compare the betting charts with the simulation to see the difference. I've notified them and they'll correct it asap.
Subscribe to:
Posts (Atom)