"build up" ?!?!

[pusha]

Hi Anastasios,

first thanks a lot for those new features, love them and works so far better than i expected. I've just got the new PreRelease and tested the new Panorama mode "build up". I've exported the chosen GAT details to excel, but i really don't understand it.

Why are the hits sum up in columns? Let say i want a prediction for my chosen pool of numbers, but the order from #1-#50 doesn't match the one in the GAT, the hit are sum up vertical, i though that the hits should be analyzed specifically for my numbers? The thing is that all 50 numbers are listed but not in the order like it is shown in GAT, can you have a more detailed description of the build up function and how it should be read and understand when exported to excel?

Is it possible to ad a chart like the Hits Delay, but instead with the order of hits in a range of draws which i can choose by a dropdown menue for like 1-xx draws?

Thanks a lot in advance a merry x-mas.


Cheers,

pusha.

[lottoarchitect]

Hi pusha,

I suppose you briefly understand the logic of build-up. In short, it tries to exploit the idea that a GAT table can sustain good hits based on the signature detected. It is just another way to look for the best GATs. The other logic available is of course the default augmentative/absolute which looks on the most hits overall.
In more detail, the principle of each GAT table is that if it can detect a good signature, a sequence of good hits will show up. The build up approach defines a way to find those GATs that are possibly in such a good hit sequence and based on that, our expectation it will deliver again a similar or better hit at the very next draw. So, if e.g. a GAT demonstrates at its last part of tested draws steady/increasing hits, this illustrates ability to maintain its hit ability (a good signature). Also the approach used in the current build up implementation is to prefer GATs that have the longest such sequence and also show a smooth improvement over the hits. That means a sequence 1, 1, 1, 2, 2, 4 is worse to 1, 1, 1, 2, 3, 4 because it gives a sudden jump (2 to 4 at its last 2 hits) compared to the latter (3 to 4). This could mean volatile signature. Of course this doesn't mean it isn't a good GAT (it may well produce a good hit at the next draw) but a very good signature should smoothly build up the hits (and the ideal would be of course to predict all the correct numbers a few times in a row). I don't say this is the best way to approach the build up criterion of signatures but we need a way to distinguish the GATs so to be placed in the panorama and I opted for this approach. I may alter that in the future if I find something better to look for. The GATs in build up are positioned in the columns based on what is the last hit achievement, so if a GAT managed 3 correct numbers at its last tested draw, it will show up at column 3.

Now, the export to EXCEL feature just exports the hidden predictions demonstrated by the red line in the graph. Each GAT does not really predict individual numbers. It predicts a string where then this is translated to numbers and the most dominant should show up at the initial positions. So, when you observe a GAT, you always have available the full string of prediction and when you adjust the req.numbers of that GAT table you just "hide" the remaining string but it is always available because the whole thing is evaluated.
Now, in the excel the first line shows some information, which is the fields and the parameters used. The GAT table the data refer to is the actual sheet (you'll see the GAT ID number at the bottom - you'll see more GAT IDs there if you export more than one GAT from the panorama). The numbers #01 #02 etc till the last one of the first line just demonstrate the order of preference decided by each GAT. So the column #01 contains the number that fall at position #1 of the relevant prediction string of each tested draw. The rest of the lines contain the actual tested draw on the left column followed by the prediction string where each #X column shows the number picked at the position #X of the relevant prediction string for that tested draw. The export does not show the actual predicted numbers shown at the predicted numbers panel in GAT (the new draw to come). It exports only the tested draws internal predictions which you can't access in previous GAT versions. Also, when you set a given req.number X at a particular GAT table and you observe its red line, the equivalent in the excel is to check the predicted numbers #1 to #X. If the red graph e.g. says 3 correct predicted numbers at the last tested draw, you should see 3 numbers marked green within the initial #X positions for that tested draw. The same way works for all the other tested draws.

Why are the hits sum up in columns?

This is to demonstrate concentration of where the most correct predicted numbers show up. A GAT that does better than pure luck should have more hits at the initial positions compared to any other equal range at other positions. I could eliminate that really, it didn't meant to be used somehow, it is just information that can be extracted from the data available.

The export is not specifically designed for the build up; it works the same way even if you display augmentative or absolute. It is just a better insight of what each GAT shows at the red line with the possibility to find some sort of pattern in there not possible to do within GAT itself and take advantage of. I have observed e.g. that when a particular prediction fails, it usually contains some of the correct numbers at the previous prediction. A GAT table should be able to understand that on its own but such a GAT will probably show up at later stages.

Is it possible to ad a chart like the Hits Delay, but instead with the order of hits in a range of draws which i can choose by a dropdown menue for like 1-xx draws?

Do you refer to actual GAT program or excel? Can you elaborate a bit more on what you are after? Adding features just for the sake of it isn't productive and I'd hate to mess the code to support that.

cheers
lottoarchitect

[pusha]

Hi Anastasios,

i will give myself more time to fully understand the details you have posted. But i'd like to answear with my thoughts on it for now.

so if a GAT managed 3 correct numbers at its last tested draw, it will show up at column 3

Which column 3 do you refer to? In the Panorama or the GAT Table itself?

So the column #01 contains the number that fall at position #1 of the relevant prediction string of each tested draw.

So the number in the column #1 (Excel Sheet) is just the number GAT choose to start the string with and not like the 1.st number shown in the GAT Table for it's favourable position itself?


It's quiet interesting to see the hits in the Excel sheets, but to me it is more important to see how the GAT numbers performed in the past draws instead of the whole GAT String because i might play with those numbers instead of those of the string right?

Do you refer to actual GAT program or excel?

If chosen build up view mode in Panorama the GAT with the highest hits in the last draw & overall will be on the top line right? But it is quiet impractical to see how the hit cycle of that GAT really is, therefor i have to see string the excel export right?

What i'd like to see is a chart like or just simple table of the past hits with the hit reached by the pool of numbers predicted by GAT. E.g. if i choose from a dropdown menu past draws are 10 than the table or chart will show the hits of those past 10 draws. If i understood correctly the Hits chart represent the hit produced by the string numbers (in Excel sheet) and not those shown in the GAT window right? But what is more important to me is that how my real numbers performed right?

One thing that came to my mind is that what i propose is based on my game style, i like to choose a pool and play it for the next 3-5 future draws to give the GAT time to hit, not like some who play with run factors, which produce less hits in MY backtesting. Maybe working with run factors just need more time to hit or simply is worse . I think if you work with run factor than the Hits performance in the prediction string itself is more important right?

Still what would be interesting is that i can extract the prediction string of a GAT naked, what i mean by naked is only the prediction string instead the ones from the past draws, maybe if you would mind to add a line above the last draw and place the naked prediction at that line, than i could see which #X column of the String is getting worse or better.


Cheers,

pusha.

[lottoarchitect]

Which column 3 do you refer to? In the Panorama or the GAT Table itself?

The panorama.

So the number in the column #1 (Excel Sheet) is just the number GAT choose to start the string with and not like the 1.st number shown in the GAT Table for it's favourable position itself?

Lets clarify first what I mean with the term "prediction string". GATs do not pick individually numbers. Each GAT table actually predicts a "complete string" which is presented to user with the predicted numbers. When you request from a GAT table e.g. 10 numbers, it just displays the initial 10 positions of that string with whatever numbers are assigned to these positions. The string is fully computed (all the positions) and contains all the numbers of your lotto game obviously. The string itself however contains the concept of where the best numbers are meant to show up and these are at the initial part of that string. To make it easier, forget about the term "string". Just consider each GAT shows you its preference of numbers in the order it believes they will show up (obviously it expects the numbers that will actually show up are at the initial positions).
Now, in excel you see only the predictions of that GAT table for the tested draws, each line representing the prediction made for the tested draw in the same line. Consider you have 1000 draws in your history and Run factor = 1. You run GAT and see its prediction within GAT. This prediction is for the next draw to come which will be for the draw #1001 when it comes available by the commission. You also export to excel that GAT table. The top draw in excel is actually the prediction made by GAT as if we had run factor = 0. So, if you actually remove draw #1000 from your history and obviously reduce run factor by 1 (make it 0 now) and make a new run, you'll see in GAT's predicted numbers the prediction shown at the excel in that line. The #X columns are just the order of preference in exactly the same way they show up when you increase the req.numbers to observe the predicted numbers in a GAT table. Position #1 is assumed to be the most favorable number picked by GAT for each of the tested draws. Position #2 is the second best etc till all the available numbers are used.
On the opposite side, assume you add the new draw #1001 and set now run factor = 2 and run GAT to get the new prediction. Also you export that GAT to excel. You'll see at its latest tested draw (which is draw #1001 now) the numbers picked by GAT (in the same order) when you run the engine with run factor = 1.
Put it in a different manner, if you show a GAT table and request 10 numbers only, you see there only the initial 10 best numbers picked by GAT. This is equivalent as if you go to the excel and delete all the columns #11 to the end. You'll just see the predictions performed by that GAT table for all the tested draws when requesting 10 numbers. I really hope you understand what the excel export is about.

It's quiet interesting to see the hits in the Excel sheets, but to me it is more important to see how the GAT numbers performed in the past draws instead of the whole GAT String because i might play with those numbers instead of those of the string right?

Not sure what you mean here but the excel shows exactly that "how the GAT numbers performed in the past draws" by the specific prediction made for each of the tested draws. Not in the sense of a powerset however if you mean that. That concept of powerset is irrelevant to GAT. I know you compare a prediction to past draws but this has nothing to do with GAT in first place. Powersets are a feature for statistical programs, does not fit in GAT.

If chosen build up view mode in Panorama the GAT with the highest hits in the last draw & overall will be on the top line right?

No, the build up looks for the longest sequence of steady/increasing hits to resemble the idea that a good signature keeps its accuracy over a range of draws. The GAT that shows up at top has the longest such sequence. It is also positioned in a column of the panorama based on how many hits it had at its last tested draw. So, all the GATs shown under e.g. column 3 had at the last tested draw 3 correct numbers predicted and the GAT that shows at the top has the longest such sequence of steady/increasing hits production.

But it is quiet impractical to see how the hit cycle of that GAT really is, therefor i have to see string the excel export right?

Based on what the build up shows, no we don't care at all about cycles anymore. We just use GATs shown in this mode for the next draw to come only. The idea is really, if a GAT can sustain such a steady/increasing hits production over several draws, it may well do that again at least for the next draw to come. No cycles are involved here. The idea of cycles (which are present in any GAT) is important to the GATs shown at the augmentative/absolute because in that case we care about its overall hits appearance over the whole tested draws and there cycles are important to observe. The case of build up GATs do not care about that because what we look for is what the GAT did over its last part and specifically GATs that show a steady/increasing hits production, thus no cycles are observable here anyway (and we shouldn't really look for them).

What i'd like to see is a chart like or just simple table of the past hits with the hit reached by the pool of numbers predicted by GAT. E.g. if i choose from a dropdown menu past draws are 10 than the table or chart will show the hits of those past 10 draws.

That is also what the excel shows. If you want only the last 10 draws, then delete all the lines below the 10th tested draw.

If i understood correctly the Hits chart represent the hit produced by the string numbers (in Excel sheet) and not those shown in the GAT window right?

Yes. It is exactly that.

But what is more important to me is that how my real numbers performed right?

How can we know that if we don't know which are the actual drawn numbers?

lottoarchitect

[winaheap]

Hey Pusha,
I had an argument in an private discussion with anastasios a while back, regarding static number sets.He just dose`nt get it.The numbers predicted by gat are at this very week a static set, which you need a line graph to visualise the actual trend of any set of numbers that gat predicts.You can see by the trend or the actual pattern of the line graph to help you decide if gat did a good job of predicting a given set of numbers.We all have different techniques of playing, i am interested in the concept of gat, but i have my own way of predicting numbers which i devised from a certain software,so until anastasios adds this line graph , i have no interest in gat.

[lottoarchitect]

Everyone is entitled to believe what approach is better to follow. Full understanding of what GAT engine delivers as output would really make some assumptions change. Indeed I completely disqualify the use of "powersets" and their trend over the past draws as a tool to evaluate the quality of the numbers picked by GAT. I'll get once more into trouble to explain why "powersets" are not good for use with GAT.

This is an extract of the last 9 predictions made by a GAT table (requesting 9 numbers) to predict The Australian Super 7s Oz Lotto game (7/45 + 2BB) including the bonus balls. The actual draw on 6/12/11 was 02 08 12 24 29 30 33 + 11 17

I picked GAT 11570 just to illustrate what I mean. Its last 9 predictions were

Code: Select all

Draw #320 - 29/11/2011   05 06 09 12 15 33 34 | 01 04-	34	08	11	04	25	39	01	21	15	(2 + 2)
Draw #319 - 22/11/2011   05 10 11 16 21 25 37 | 12 39-	28	11	10	17	20	39	15	16	27	(3 + 1)
Draw #318 - 15/11/2011   06 07 08 09 20 22 28 | 29 39-	39	35	27	14	22	44	23	29	08	(2 + 2)
Draw #317 - 08/11/2011   01 03 06 15 18 21 34 | 12 28-	32	14	16	07	08	20	05	39	30	(0)
Draw #316 - 01/11/2011   05 12 17 23 25 31 44 | 13 19-	28	36	21	30	33	24	08	07	39	(0)
Draw #315 - 25/10/2011   03 16 29 33 34 35 36 | 14 31-	21	03	44	45	39	05	33	25	35	(3 + 0)
Draw #314 - 18/10/2011   06 10 15 16 36 43 45 | 09 44-	04	05	02	16	08	12	39	34	21	(1 + 0)
Draw #313 - 11/10/2011   06 08 12 25 32 39 44 | 10 41-	09	02	27	43	21	26	11	28	20	(0)
Draw #312 - 04/10/2011   01 02 09 19 25 36 42 | 05 06-	08	42	30	45	40	02	33	35	29	(2 + 0)

Also here is a graph illustrating the hits achievement of that GAT table in predicting the tested draws.



Uploaded with ImageShack.us

The last 9 red dots are actually the hits illustrated for the last 9 tested draws. You see there this GAT over the last 3 tested draws constantly producing a 4 match. You also see the dynamic nature of GAT's prediction that drastically changes from draw to draw. The actual prediction for draw 6/12/11 was

Code: Select all

Draw #321 - 06/12/2011 02 08 12 24 29 30 33 | 11 17-	42 44 09 19 12 30 34 11 24	(3 + 1)

You see clearly that the trend of 3 consecutive 4-hits continues. This is really a demonstration that signatures detected by GAT can keep up their prediction accuracy. Now, if you actually use that predicted set to make a comparison versus all the history (or those 100 last tested as in my example here), you just get an average estimation on the quality of the numbers long term, nothing to do with what happens at the very next draw. Which graph would you trust really?
What is the correlation to the actual trend seeking GAT does here? The trend is obvious as you can see, 3 times 4-hits sequentially and this continues. If you use that set as powerset, you'll certainly find no correlation. The only possible good use a powerset could have in GAT is to verify that IT WILL NOT PICK any prediction that has already occurred in the past. This doesn't happen anyway so even that is pointless to do in GAT to verify it.

If you actually compare the prediction 42 44 09 19 12 30 34 11 24 as powerset you have

Code: Select all

Draw #320 - 29/11/2011   05 06 09 12 15 33 34 | 01 04-   (3 + 0)
Draw #319 - 22/11/2011   05 10 11 16 21 25 37 | 12 39-   (1 + 1)
Draw #318 - 15/11/2011   06 07 08 09 20 22 28 | 29 39-   (1 + 0)
Draw #317 - 08/11/2011   01 03 06 15 18 21 34 | 12 28-   (1 + 1)
Draw #316 - 01/11/2011   05 12 17 23 25 31 44 | 13 19-   (2 + 1)
Draw #315 - 25/10/2011   03 16 29 33 34 35 36 | 14 31-   (1 + 0)
Draw #314 - 18/10/2011   06 10 15 16 36 43 45 | 09 44-   (0 + 2)
Draw #313 - 11/10/2011   06 08 12 25 32 39 44 | 10 41-   (2 + 0)
Draw #312 - 04/10/2011   01 02 09 19 25 36 42 | 05 06-   (2 + 0)

What possible conclusion do you make out of this anyway about the performance of that powerset??? The only thing I can gather is that it potentially may contain 1-3 correct numbers every so often. I'm sure you'd drop that prediction, wouldn't you? The case is really, if you take any 9 numbers and compare them like that, you'll get a similar outcome more or less, nothing to do with actual trend detected by GAT's signature. We simply can't judge the quality of a set of numbers as powersets.

You can see by the trend or the actual pattern of the line graph to help you decide if gat did a good job of predicting a given set of numbers.

I really hope you understand by the above that this sentence does not hold true. The best line graph for trends seeking of GAT is the red line already shown.

cheers
lottoarchitect

[winaheap]

Hello Anastasios,
Like i said, we all have our own methods of playing,the reason the line graph of the predicted set from gat is valuable ,i mean the powerset, is for example, if you are using gat to predict, let`s say 6 numbers,we use the powerset graph to determine if at LEAST one number will appear as gat has predicted.My strategy is to try and predict 3-4 unique sets of 5-6 numbers with each group containing at least 1 number.I use the powerset graph from years of experience,in determing certain patterns that eventuate,ex : when a certain pattern appears what happens most of the time in the past,so what happens in the past IS most likely to happen in the future.This way i determine with high level of confidence if what gat is predicting(at least 1 number to appear) is most likely with the help of powerset graph to eventuate.I still don`t get your stubborness on this issue.
It`s quite simple really, anything that can help us to improve our confidence in prediction should be utilised,i for one can`t see the problem here,or is it because you are a scientist, you disregard any form of "magic" you must work strictly to the scientific rule? If so let me remind you, we are dealing with lotto here, we all need a bit of "magic" or "luck" to help us crack this game,just take that on board please.

[lottoarchitect]

Erm, ok... the trick here is to understand that we don't deal with powersets logic in GAT but with the capability of signatures to sustain good hits over a few draws - there is no way a powerset comparison to reveal that and by the way that was the example I demonstrated above. Whatever you said above is more accurately measured by the graph shown in GAT because of that difference. Of course you can use any approach you want. Powersets comparison have nothing to do with GAT's signature logic therefore not included. It is not about being stubborn or scientific about it, it is just an approach not suitable to evaluate the performance of the sets picked by this system.