Development updates

Any latest news will be posted here...
User avatar
lottoarchitect
Site Admin
Posts: 1450
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: Development updates

Post by lottoarchitect » Fri Jun 01, 2018 1:04 pm

It goes way beyond picking just a GAT, It attempts various methodologies and tries to pick among them based on a overall range performance. It does use GATs that are meant to end up to the panorama in those methodologies.

bluemoon
Regular
Regular
Posts: 25
Joined: Sat May 05, 2018 3:47 pm

Re: Development updates

Post by bluemoon » Fri Jun 01, 2018 10:12 pm

It goes way beyond picking just a GAT, It attempts various methodologies and tries to pick among them based on a overall range performance.
It does use the best process, I like it!
It does use GATs that are meant to end up to the panorama in those methodologies.
I’m still not sure what’s the meaning above. Could you explain it a little bit detail?

Sincerely,

User avatar
lottoarchitect
Site Admin
Posts: 1450
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: Development updates

Post by lottoarchitect » Fri Jun 01, 2018 11:37 pm

It means it takes into account, for any methodology checked, the GATs that eventually can make it to the panorama and decides on how to use them, which ones to use or skip or whatever the methodology suggests. Can we please stop such detail discussion at this point? It is not finished yet. When it is ready, I'll post detail of what everything does.

bluemoon
Regular
Regular
Posts: 25
Joined: Sat May 05, 2018 3:47 pm

Re: Development updates

Post by bluemoon » Sat Jun 02, 2018 1:10 am

I’m so sorry about that!
I thought the 3.0 is almost done and my comments should be welcomed here.
Do you have an exact date (or month) of releasing 3.0 now?

Sincerely,

User avatar
lottoarchitect
Site Admin
Posts: 1450
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: Development updates

Post by lottoarchitect » Sat Jun 02, 2018 3:34 pm

Not yet, it will be announced here.

bluemoon
Regular
Regular
Posts: 25
Joined: Sat May 05, 2018 3:47 pm

Re: Development updates

Post by bluemoon » Mon Jun 11, 2018 3:09 am

ONE SUGGESTION

Dear Mr. Lottoarchitect:

According to your introduction about the both modes Auto & Expert in GAT 3.0, I have one suggestion as following:

1) Could you partially transplant the algorithm or technology about “the best process” in Auto mode into the Expert mode of GAT 3.0? i.e. “tries to identify the best overall picking methods to use for selecting out some better GATs from the all displayed GATs”.

If so, users would have two options: to get directly the “best” numbers from Auto mode or to get directly some better GATs from Expert mode, then to select the best GAT from the fewer better GATs which are generated by “the best process ” in Auto mode by them-selves similar to use 2.3b.

2) In this way, there would be two new modes in 3.0: Brand new Auto mode and improved or semiautomatic Expert mode. So, I suggest that you can double charge the GAT 3.0 at its release since the GAT 3.0 has been a complex of two new versions; especially the Expert mode has been improved very well comparing the old Expert mode in 2.3b.

How about your opinions?

Sincerely,

User avatar
lottoarchitect
Site Admin
Posts: 1450
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: Development updates

Post by lottoarchitect » Mon Jun 11, 2018 4:42 pm

This isn't applicable because of the way automatic is meant to operate. No single GAT is picked, if it did I would allow in manual mode the ability to flag the individual picked GAT, but this is far far away from what automatic does underneath. I'm not even sure if a user can approach what automatic does using the manual/expert mode. The only connection of automatic to the manual/expert mode is automatic uses the basic building blocks which are the GAT tables that end up to the panorama, and even this is not entirely true. What I can suggest is to wait till you get this system, then we can discuss further improvements. Right now the design has a specific implementation in mind since its inception but it is not limited to this implementation. Automatic is a whole program within the program, so expansions to its operation are expected too in releases beyond 3.0.

bluemoon
Regular
Regular
Posts: 25
Joined: Sat May 05, 2018 3:47 pm

Re: Development updates

Post by bluemoon » Tue Jun 12, 2018 2:37 am

Thanks for your detail reply!
So sorry for my misunderstanding to GAT Engine 3.0.

Sincerely,

bluemoon
Regular
Regular
Posts: 25
Joined: Sat May 05, 2018 3:47 pm

Re: Development updates

Post by bluemoon » Tue Jun 19, 2018 3:47 am

Still the default 20 stat.data is a good setting for all situations but if the automatic stat.data evaluation picks another value, I would prefer that instead for the slight boost it will provide over the default setting. In my test runs it has even picked values like 120 or 175 and when compared to the standard 20 setting, indeed I see some better hits production at the various panorama displays (tested on Australia's Monday Lotto mostly) which ultimately means more chances to pick even the one GAT that will produce the good hit at the next draw (this regardless the RF utilization which is the primary usage).
Mr. Lottoarchitect:

I’m still not sure what your meaning above is.
The default stat.data is 100/20 but now your setting is 120 or 175.
Does the 120 (175) of stat.data refer to 100/120 (100/175) or 120/120 (175/175) or others? Australia’s Monday Lotto is a game of Pick-6. Does this setting (120 or 175) also apply to Pick-5? If not, what settings would be available for Pick-5? OR, do you have a formula/strategy to select the setting of stat.data for the different games?

Also, could you explain in detail about “indeed I see some better hits production at the various panorama displays” when you’re using the setting of 120 or 175? Let’s say, did you find that there is a high value of Imp-R or Ratio % under this setting? OR are there outstanding performances in Blue and Red lines? (Assume that the above situations occur in Augmentative mode only. But maybe the Build up mode is working for the settings too, I’m not sure. )…. If possible, could you display it by using some pictures? Thanks.

Sincerely,

User avatar
lottoarchitect
Site Admin
Posts: 1450
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: Development updates

Post by lottoarchitect » Thu Jun 28, 2018 12:06 pm

I refer only to the stat.data value which by default is set to 20. This value is a balance between computation speed and enough data to be used for each individual prediction performed to generate the GAT's history performance and the future unknown draws. Now, the automatic stat.data evaluation is computed based on the idea that over a lengthy test or GATs, it picks the stat.data value which maximizes hits production among all those tested GATs, so e.g. if stat.data 10 can generate let's say 10000 hits over the course of the initial 1000 tested GATs and stat.data 122 can generate 11000 total hits over its (equivalent in length) relevant initial 1000 GATs, then stat.data 122 delivers slightly better performance compared to stat.data 10. The differences are not huge really but we always favour the results that give some slight boost and in this case the criterion is always more hits production.
I mentioned stat.data 120 or 175 as an example result I got here, that means the system detected that these particular settings for that particular lottery (120 picked by the system for lottery A, 175 picked by the system for lottery B) for that particular time period analyzed those stat.data settings did deliver more hits overall over the initial X GATs tables, which ultimately means a slightly better focus of the whole propagation. This is what the automatic stat.data evaluation does. Again, the default setting of stat.data 20 is perfectly adequate for all runs, the real benefit of increasing stat.data is only in games that have a very obvious - larger than 20 draws - cycle in the data in which case the stat.data value should be greater than the larger observable cycle of the data. Don't worry too much about this, the bottom line is we should always use automatic stat.data evaluation to compute a new value every so often but the effect is not so big anyway. The real benefit of this automatic evaluation is we surely can avoid very bad settings for stat.data. For GAT 2.3 and even the new 3.0, stat.data 20 is perfectly fine as cycles are not really observable anyway in real lotteries, however since we can now auto-evaluate this in the new version, no point to stick at the default 20 if a slightly better setting can be detected.

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest