Development - G.A.T. Engine 2.3

Any latest news will be posted here...
Post Reply
User avatar
lottoarchitect
Site Admin
Posts: 1635
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Development - G.A.T. Engine 2.3

Post by lottoarchitect » Tue Nov 27, 2012 10:54 am

GAT's development is well underway. The version is by no means complete since there are a lot to be added. So far the following are in place:

- Delay graph to show a line of actual delay data (requested feature). This shows the actual delays in a graph for a given hit to occur. So now we have a summary (current version) and a line showing the actual delays. To access this feature, we must observe a specific hits delay (i.e. delay of 3 hits) and it is available by right-clicking the delay graph. The most recent delay (not the current delay shown at the bottom) is the right-most point.

- Copy the predicted numbers in sorted order (requested feature). The "copy to clipboard" function currently copies the predicted numbers as shown (unsorted). SInce it is easier to see what numbers are in there when sorted, the copy button now copies the numbers sorted and if someone needs the order of preference decided by GAT, right-clicks on the copy button for this function.

- Colouring GAT cells in the panorama to compare their hits to a given set of numbers. This can be used to instantly show which GATs managed to predict the actual draw (draw must be removed from the history or not adding the new actual draw for this comparison to make sense) and the darker the colour, the more numbers actually predicted. Dark green means more numbers correct, light green means fewer. This function needs further improvement e.g. it will be possibly moved to the functions toolbox but it is functional as it currently is. This feature is enabled if we enter any number for comparison at the options. Also, when the comparison is active, additionally the predicted numbers of each GAT are colored green to indicate which numbers have been predicted (that was a feature of GAT 2.1). The colors used are spread from 0 to the minimum of current requested numbers and total numbers to compare. So if e.g. we enter 5 numbers to compare and show req.numbers 3 at the panorama, obviously the maximum possible is to match only 3 numbers from out set. Therefore the dark green will show up when 3 matches occurred (the maximum) and the other cases will gradually fade to light green (the default color used in panorama) which indicates 0 hits. Similarly for the same 5 set of numbers, if we display req. numbers 20 at the panorama, the maximum possible will be 5 matches (all the numbers in the compare set), therefore dark green is assigned to any GATs that will match all 5 numbers and gradually fade to light green (the default color of the panorama) if 0 hits achieved when compared to that set. This feature is really an eye-opener! Also makes a very nice visual mosaic :lol:

- Max GATs shown per column (up to 999). This is another speedup mechanism. This actually will be used to limit the need to keep track of more GATs in a given column since most good GATs are expected to show up at the initial few top cells shown. Even having 99 GATs to check in a column is already too much. Not sure if anybody will set this to a higher setting but it is available if he wants to do that.

- X, X-1, X-2 feature. This is a very important addition. It is fully operational but still needs work to be faster in operation and it is recommended to use the Max GATs per column along with the req.numbers/category speedups. Currently the panorama shows only the top X hitters. However they can be in a cold cycle and since fewer and fewer make it to the panorama as we progress the scan there are more chances to pick a cold performing GAT. This feature is selectable (enable/disable). If disabled, the engine works as it currently does. If enabled, it displays all the top performing GATs in terms of hits produced in descending order till all cells in the column are filled up. So e.g. the initial top cells will have GATs producing X hits, the following ones X-1 hits etc till all the cells are filled up. This feature used together with the compare numbers feature show right away where the best GATs show up where the best ones are expected to be close to the top, if they are not the top X hitters due to their cold cycle. A selection of a good GAT is expected to occur within the X, X-1, X-2 hitters anyway (the whole point of this feature really!). Also, when browsing a column with this feature enabled, the order is preserved so we'll have the top hits performing GATs shown first.

- Directly export predictions from the panorama to the clipboard. The selection box will be used to identify which GATs we want to export the predictions and right-clicking in the panorama will bring up the function.

Prerelease versions will become available as new features are added since the features list to be added is huge. Please keep in mind, prereleases are not fully verified for correctness (bugs might slip in) but given the official version with all the added features will take time, it really makes no sense to delay features that are already implemented, even if they can take further improvement later. The GUI might look a bit mess (i.e. boxes and options or even functionality might move around till the final official release) due to development.

So, if there is demand for this current prerelease I can prepare one for download at the account management. The point is to use and comment on existing added features along with other proposals till the final version is formed in due time. Finally, since there is no help file yet for these new features, any usage questions should be posted here.

cheers
lottoarchitect

Skirrow
Advanced
Advanced
Posts: 53
Joined: Tue Jan 29, 2008 8:23 pm
Location: UK

Re: Development - G.A.T. Engine 2.3

Post by Skirrow » Tue Nov 27, 2012 10:52 pm

Hi LA ,
Pleased to hear the next version of GAT is under way ,much appreciated .

I would like to play around with the pre-relaese version as soon as you make it available ...

Thanks

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

Re: Development - G.A.T. Engine 2.3

Post by lottoarchitect » Wed Nov 28, 2012 4:25 pm

The prerelease is available at the account management. The zip file contains only the executable, place it at the GAT's installation folder, make a shortcut to desktop and also enable the compatibility/run as administrator before running the prerelease (if needed - check the original GAT icon to determine that) to avoid license locks. If you have questions or comments post them here. The evaluation engine is the same to 2.2 so you can directly compare the prerelease to the current version.

cheers
lottoarchitect

charliez
Casual
Casual
Posts: 15
Joined: Wed Oct 17, 2012 6:01 am

Re: Development - G.A.T. Engine 2.3

Post by charliez » Tue Dec 04, 2012 12:14 pm

Hi LA how do I get to the account manger to get the new ver of gat pls
regards
charlie

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

Re: Development - G.A.T. Engine 2.3

Post by lottoarchitect » Tue Dec 04, 2012 2:57 pm

Go to the main webpage of GAT and login with your user ID and auth.code. The download is available at the same page you downloaded the license file.

lottoarchitect

baalhabait
Advanced
Advanced
Posts: 92
Joined: Wed Apr 11, 2012 9:58 pm

Re: Development - G.A.T. Engine 2.3

Post by baalhabait » Fri Dec 07, 2012 6:52 am

Hits indeed occur within the x-1, x-2 hitters, but the real issue is what about the high hits...You know, hits that will deliver a little...just a little more than half of the investment in winnings.
Those who cannot afford playing 800 blocks per draw will find out all those 5 and 6 hitters occurs mostly in x-10, x-20 area - a hopeless area.

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

Re: Development - G.A.T. Engine 2.3

Post by lottoarchitect » Fri Dec 07, 2012 9:14 am

Well, not sure if I should consider this a positive or negative comment. First, how many numbers do you request? Second, what are your expectations regarding 5/6 hits? Have you computed the chances to randomly match that amount of hits and compared to what GAT delivers is hopeless? I actually see very favorable results here. For example in a 6/49 game (that's the one I currently test here now) the chance to pick 10 numbers and have 6 correct in there is 1 in 66589. That means I should present 66589 predictions to find one there that delivers just one 6 hit. Now, given that you cannot display more than 999 GATs per column and you have 6 columns shown there, I present 999 * 6 = 5994 10 number predictions. That means the engine already delivers 66589/5994 = 11,10 times better than pure luck! And if you can actually find that 6-hitter in a column constantly (i.e. the 6 hit column) this goes up to 66,6 times better than luck. So, even if you pick randomly any of those 5994 predictions, you are still at least 11 times better that playing a random selection picked by hand. That sounds to me much less hopeless really. Furthermore, I have tested numerous runs for predictability on lower hits (3/4/5) and even when I display only 19 GATs per column, I get more than double the amount of 3 hits normally expected (that happens in almost any column I test). Not only that but I also find in the same column, those 3 hitters to be actually 4 and even 5 hitters. So, if I pick randomly (when picking 10 numbers) any GAT in that column (not caring about its hit graph at all - no investigation on what GAT to pick), I see in there 4-6 3+ hitters. Just to compare it to natural chance, in 19 predictions of 10 numbers, I would normally expect to have at most 2 3-hitters and none of 4/5 hitters. But I actually get 3-4 3-hitters, 1-2 4-hitters and 0 or 1 5-hitter. And that within 19 10-number predictions! So really, how can you say this is hopeless? What I see here is very good predictions we can't even dream having by other means; all that happens with 2M GATs scanned. To be fair, I get there results in the development prerelease 2 where I have changed the GATs ordering to take into account the overall hits production in all hits categories besides the hits achieved in hit category X (current prerelease 1 doesn't care for that). But still I wouldn't expect much different results with prerelease 1.

lottoarchitect

baalhabait
Advanced
Advanced
Posts: 92
Joined: Wed Apr 11, 2012 9:58 pm

Re: Development - G.A.T. Engine 2.3

Post by baalhabait » Fri Dec 07, 2012 9:46 pm

Of course there is very favorable results here! GAT is awesome compared to random, Without using GAT i would totaly lose my belief in this - but given we're using the bare minimum - week trends that suggests eg. some 3hits with one 5hit or one 6hit are existed in massive amount indeed - i would expect this to happen more often also with more obvious 3hit trends.
Apart of that, I think it would be great if a slider could be added for each GAT hits graph, so the hit observation will be shown more clearly, especially when the tested draws goes beyond 150.
Currently the more the amount of tested draws is used...the more the hit graph of each GAT becomes messy because all the hits are forced to be shown at once.

baalhabait
Advanced
Advanced
Posts: 92
Joined: Wed Apr 11, 2012 9:58 pm

Re: Development - G.A.T. Engine 2.3

Post by baalhabait » Sat Dec 08, 2012 1:44 am

Of course there is very favorable results here! GAT is awesome compared to random, Without using GAT i would totaly lose my belief in this - but given we're using the bare minimum - weak trends that suggests eg. some 3hits with one 5hit or one 6hit are existed in massive amount indeed - i would expect this to happen more often also with more obvious 3hit trends.
Apart of that, I think it would be great if a slider could be added for each GAT hits graph, so the hit observation will be shown more clearly, especially when the tested draws goes beyond 150.
Currently the more the amount of tested draws is used...the more the hit graph of each GAT becomes messy because all the hits are forced to be shown at once.
BTW, Can you tell how long it take you to compute 2M GATs?

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

Re: Development - G.A.T. Engine 2.3

Post by lottoarchitect » Sun Dec 09, 2012 9:30 am

I kept a note on the graph bar for inclusion. I currently use a third party graph which is good but doesn't offer much flexibility. Either I have to develop my own graph system (which is in plans but not right now) or find a way to trick the current graph to do this. As for the time needed, it was a few hours (didn't really counted them) probably around 5-6 hours but I also used the speedup 1-15 req.numbers. However I have a quite fast computer here.

lottoarchitect

Post Reply

Who is online

Users browsing this forum: No registered users and 17 guests