New GAT 2.2 coming soon!

G.A.T. Engine general discussion
User avatar
Mindas
Casual
Casual
Posts: 13
Joined: Thu Mar 05, 2020 10:09 am

Re: New GAT 2.2 coming soon!

Post by Mindas » Sun Mar 15, 2020 12:08 pm

lottoarchitect wrote:
Thu Feb 09, 2012 2:56 pm
My experience with this version is that we can find quite better performing GATs the more we let it run. An 24 hours run will yield possibly the "almost best" that can be found and a 6 hours run will definitely give better performance than the current 2.1 version can achieve.
Anastasios, can you please suggest optimum time run specs for version 3.0.2?

User avatar
lottoarchitect
Site Admin
Posts: 1635
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: New GAT 2.2 coming soon!

Post by lottoarchitect » Sun Mar 15, 2020 6:17 pm

There isn't any optimal run-time, the more you let the engine run the more chances to trap an even better overall performing GAT to use. There is some slight speed increase in GAT 3.0 vs older versions due to internal improvements but the same advice still apply. Most runs are in the range of a couple to 5-6 hours and for all practical reasons 24 hours would be the most I would let the engine scan for. Keep in mind, to continue use with Run Factor, that same amount of time plus a bit more will be needed the next time, so in reality GAT scan times are completely up to the user to decide; how much time he has available to run a scan for. I can say reaching up to 2 million GATs is a good target to be but even if we stop the scan at half a million, still worthy GATs can be found in there to pick and keep using.

User avatar
lottoarchitect
Site Admin
Posts: 1635
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: New GAT 2.2 coming soon!

Post by lottoarchitect » Sun Mar 15, 2020 6:36 pm

Mindas wrote:
Sun Mar 15, 2020 11:52 am
lottoarchitect wrote:
Tue Feb 07, 2012 9:11 am
The very first GAT #1 computed has a very basic such analysis of randomness. It passes its own understanding of randomness to GAT#2 which refines this randomness further. This goes on and on for as long as we want it to. So each GAT X actually represents a set of signatures as evolved and influenced by all GATs #1-#X-1. So in order to compute GAT #1000, we need the information of GATs 1-999. During that process, GATs emerge that are better performing than the previous GATs. This represents a GAT that understood better that inherent randomness and in effect it provides more overall hits and also you may notice successive hits at that GAT's graph (red line or blue line going up).
So, if I understand correctly, each GAT is a "better" version of the previous one in terms of next draw prediction.
Given that, shouldn't the order of GAT IDs that make it to the panorama, always be in descending order in the same category and in ascending order from smaller category to a greater one?
For example if GAT #457 made it to the 1st place of category 1, then every GAT placed in category 2 should have greater ID (>457) and on the opposite, the GAT in the 6th place of cat 1, should have ID<457?
And finally, if this is correct, a major factor in getting the best performing GAT, wouldn't it be the run time we give to the engine?
Given that, shouldn't the order of GAT IDs that make it to the panorama, always be in descending order in the same category and in ascending order from smaller category to a greater one?
Yes, for as long as we talk about the same x-hits performance. GATs in each category are placed based on the X, X-1, X-2 rule. For example, assume we check the 3 category GATs. The top GAT delivers let's say 50 3-hits and there are also another 4 GATs also delivering 50 3-hits. Apart from that basic distinction, those 5 GATs are placed in the top 5 positions of the 3 category based on some extra analysis (specifically the Hits+delay) internal analysis. Most of the time this translates as higher IDs are positioned above lower IDs but this is not always the case. So, these 5 top GATs will be placed in order based on their hits+delay performance since their initial criterion of 50 3-hits is the same. To continue this example, any GATs that produce 49 3-hits will be placed below those 5 GATs etc. This is the X, X-1, X-2 etc positioning enhanced with hits+delay analysis.
For example if GAT #457 made it to the 1st place of category 1, then every GAT placed in category 2 should have greater ID (>457) and on the opposite, the GAT in the 6th place of cat 1, should have ID<457?
It boils down to what each category represents. The idea of "category" is to present the best performers in a particular hit category i.e. 3-hitters if we talk for category 3. This DOES NOT mean that e.g. the top 2-hitter (which shows at the top of 2 category) is also a top 3 hitter. So, to answer that question: NO. To be precise, this refers to the augmentative/absolute panorama modes. The ID is just an identification, suggesting in which order a GAT has been produced. What matters is what that GAT carry in terms of performance and certainly any ID that comes after another ID does not mean it is better in all aspects; in fact is can be worse.
And finally, if this is correct, a major factor in getting the best performing GAT, wouldn't it be the run time we give to the engine?
Yes, you give the engine the time to explore more and more possibilities of volatile dynamics hoping to trap even better whatever is hidden in the results. The more the engine runs, the more chances to find even better overall performers (GATs) to pick from. Please note the word "better overall". GAT's primary operation is to use it with Run Factor, meaning you find a good performing GAT and keep using it in the future draws till it delivers what it promises and the expectation is you'll get that hit quite sooner than naturally expected. This does not mean that if you scan for 5 days, the GATs that will end up in that panorama will produce the good hit at the very next draw: it really means their hit characteristics are such that if you pick higher IDs and keep using them with Run Factor you'll have better chances to get the good hit even sooner (better overall hit performance) compared to running the engine for e.g. 1 hour and pick one GAT that showed up in that scan time. For all practical reasons, a few hours scan is a sensible time to scan for but the user is free to decide how much he may want to wait for results.

Post Reply

Who is online

Users browsing this forum: No registered users and 13 guests