Forester,
I have been studying roulette spin videos. Am I correct in assuming that an accurate timing of one revolution of the ball can be used to predict the time of, and distance to, the ball’s drop onto the wheel? If that is true, then an accurate timing of a wheel revolution can be used to predict the wheel number at ball drop.
anam
Until you have an answer from people much more experienced than I am, I would like to point out that the physics of roulette seems to tell us that this is the case!
The ball is pushed downwards by gravity, which is constant. It is also pushed up by centrifugal forces, which are an effect of the balls speed and which deminuate as the ball slows down due to friction and aerial resistance. As long as the speed of the ball is high enough to generate a centrifugal force which is greater than the gravity force acting on the ball, then the ball stays in its orbit towards the outer wall on the frame around the wheel. Once friction has lowered the speed under the critical (and constant) treshold where centrifugal forces are equal to gravity, then the ball starts falling down towards the wheel.
So, it does not matter how hard the ball is thrown away at the start. It will still start falling from the ball track at the same speed every time. One could plot a curve showing speed versus time which should be the same for a certain combination of wheel and ball. Then you need only to know one single point on that curve, i.e. the speed of the ball at one point in time, in order to determine the time until the ball has slowed down enough to start falling of the ball track. So, given the speed of the ball, you can find the only point on the curve which corresponds to that speed. Then evertything will follow that curve and you can see how long time to fall-down remains. Then consider how the wheel will be turned at that future point in time. Easy!
This only assumes that friction and air resistance together are constant given any speed of the ball. Air resistance should be a fairly sttrict funtion of the speed, acting more intensly when the speed is high than when it is low. But what about friction? What if the croupier spins the ball around its own axis in one direction one time, and then in another direction the other time? (I.e. once from pole to pole, once along its equator, once tumbling around etc.) It should make a difference for friction whether the ball is spinning along or against its direction of movement. However, there are reasons to believe that not even this phenomena poses any serious threat! The rotation of the ball should loose energi in the same degree as it affects the balls speed. After a few rounds, the rotation should have been largely normalized every time. That is, the speed/time curve could be slightly wider further from the fall-down point than closer to it. However, the ball automatically gets itself back on track. Every time!
The big problem is in precision of measurements. A slight error might predict that the wrong diamond gets hit. But then again, on average prediction can still be good enough to improve odds from 1/37 to 1/35 to make it profitable to play.
Good question and good answer from Rollo, All of that is valid if wheel is perfect. Slight tilt on wheel or imperfections on ball track can change it. But answer is still yes. Prediction is only possible if that statement is true. How much prediction vs. result change based on Rollos facts as air pressure or gravity is dependent on how system is based and on which way we are defining speed of the ball.
Example1
We clock 2 rotations 1 sec and 1.2 sec and ball after that travels 10 sec
Example 2
We clock 2 rotations 1 sec and 1.2 sec and ball after that travels after that 9.5 sec
In example 2 it is imposable to have ball to the end time change without change in time of clocked spins. Therefore example 2 would be more realistic as follows. Example 2, if we clock 2 rotations 1 sec and 1.25 sec and ball after that travels after that 9.5 sec. This is not exact figures but only indication that deceleration of ball must increase with reduction of ball traveling time. If that’s true, then an accurate timing of a wheel revolution can be used to predict the wheel number at ball drop point. I think you mean here, accurate timing of the ball revolution to be able to define ball drop point then accurate timing of the wheel to find out position of the wheel at that moment. Minimum of 2 ball rotations are needed for prediction. But there is much more for an accurate prediction.
Forester and Rollo thanks for the informative replies.
Do either of you think a more accurate (in terms of percentage error) ball timing could be obtained by clocking multiple revolutions rather than single ball revolutions? For example, we would time a group of 5 ball revolutions and plug that into our previously established prediction equation made from video studies of a similar wheel and ball. In studying my roulette video, the plots I obtain from groups of revolutions are ‘smoother’ than those of single revolutions. In addition, the equations derived are more accurate in terms of R-squared values. Note that I use Excel to plot my data and derive the equations.
@Forester
You state … Minimum of 2 ball rotations are needed for prediction. But there is much more for accurate prediction.
You have stated numerous times that making timings with hand or toe switches is prone to errors, especially when trying to make timings of the rapidly spinning ball. Your E1, E2, V1, and V2 methods all make use of observation of the ball passing a reference point to make the prediction. Am I right in presuming, then, that your method of making super accurate ball timings involves observing the ball pass some reference point?
Also, since the wheel spins at a constant rate, would it be possible to use it as a ball timing device?
anam
Good question and good answer from rolloThank you! I must again point out that I am new to the idea to predict roulette and that I do not have observations to confirm my reasonings. So I'm happy to see that it seems to agree with what very experienced players like Forrester have observed in real life. I welcome all critisism!
Slight tilt on wheel or imperfections on track can change it.Yes, but of course rather to further advantage of the player, given that this tilt is accounted for. As I understand it, a sufficently large tilt makes the ball accelerate on one half of the ball track (when actually moving downwards), and decelerate the more harder on the other half where the tilt makes the ball track lean upwards. That means that the ball NEVER CAN fall down on the downward leaning half of the table. It can only start falling when deceleration makes it slow down so that the centrifugal force caused by the balls speed falls below the gravitational force.
In this golden situation it is basically sufficient to predict the number of rotations the ball will make (fractions of rotations are not necessary) and how the wheel will stand at that final rotation. This I believe is possible to do without the aid of contraptions, although they would help. By betting on the half of the wheel which is going to be turned towards the upwards leaning half of the ball track during the final rotation of the ball, the odds should theoretically be improved to 1/18 (plus error in prediction of wheel position) which is enourmous, in the long run DOUBLING all the bets!
If the tilt is less than enough to actually make the ball accelerate, there will only be a tendency to the effects above, but still, not much is needed to get ahead of the 1/36 odds.
One should first make enough observations of the wheel to see whether it is tilted or not. If it is tilted, use a separate method for prediction. If it is not tilted, use a method for prediction adapted to that case.
Do either of you think a more accurate (in terms of percentage error) ball timing could be obtained by clocking multiple revolutions rather than single ball revolutions? For example, we would time a group of 5 ball revolutions and plug that into our previously established prediction equation made from video studies of a similar wheel and ball.I think that you will get better precision when measuring the ball speed over a higher number of revolutions, yes. The measuring error can be assumed to be constant in terms of centimeter (of angular degrees or pockets or whatever unit used). The more rotations this error can be spread out on, the lower it gets as a fraction. 5 rotations as compared to 1 rotation should reduce the measuring error by 80%! In principle.
However, one must consider that to measure the speed one must make measurements at two points in time and space. BUT, the speed measured is valid for the ball at only one point in time and space! One can never measure what speed the ball has NOW. One can only measure the speed which the ball HAD in between the clicks. If deceleration is at a constant rate, then the speed measured was the speed which the ball had exactly in the middle of the two measuring points. At your first click, the ball had higher speed. At your second click it had a lower speed. Half TIME in between (which is NOT exactly the same as half WAY, 2.5 rotations), the ball had the speed which you measured.
To complicate things further, the deceleration is certainly not constant, but friction and air resistance has greater effect earlier when the ball travels faster, than later when the ball is slower. Assuming a constant deceleration gives a smaller error if measuring over one single rotation, than if measuring over multiple rotations. So there is a trade off between the two errors: error of timing the clicks exactly, and the error of non-constant deceleration. Maybe 2 rotation is a good trade off? This reduces the error of timing the click with a wapping 50% and still the ball has not decelerated too much, given that it is already pretty slow so that the deceleration rate has decreased which makes deceleration more constant. Also, the speed measured would be the speed which the ball had when passing (close to) that same reference point on the rotation between the two clicks, which might simplify things (measuring over odd number of rotations, including 1 rotation, would measure the speed of the ball when it was about 180 degrees from the reference point).
The same applies to measuring the wheel speed, however, I THINK that it is practical to ignore deceleration in that case, because it is pretty small, and assume constant wheel speed from time of measurement to time of ball drop.
Anam, my visual system uses wheel as reference, but I would more define it as theory then practical system. Timing device is required.
Rollo you are very correct and there is much more to all of what you said.
E2 is very easy system and I think it is easy to understand. It has advantage because we do not measure the fastest object (the ball) we observe change. Performance is ok but not the best because the system is not all the time perfectly balanced.
Clocking ball is hard and there is much more than what usually people would think.
Deeper you go in, you find more problems. Even writing program is not simple.
You need to know how long each operation takes and take that in to account.
I am using assembler, it is the hardest program but it is the best.
Many other factors are important, related to particular wheel, wheel can look the same but deceleration can be different. I take months of work to find the best solutions.
Tilted wheel.
You are close to be correct but again there is more.
You can find some good information on GG forum discussion in between Mark and Fadil. I can see that you are analytical so if you read it you will notice that even Mark which developing tilted system for years had no idea about additional factor which can create errors from0-12 numbers dependently on deceleration of the wheel and strength of tilt. He is lucky in prediction because part of those 12 numbers is balanced with different distance from diamond until ball hits rotor. But that is not fully adjusted and it is not same for all wheel speeds.
Forester
Great site, congratulations.
I note your reference to the timing error problem inherent in most systems where data is collected manually.
My backgound is in designing ICs that detect patterns in data in real time. I am planning a project to create a hardware solution to measuiring wheel/ball timings without manual intervention - hence my visit to your site. When I’ve made some progress I’ll post news.
I saw a tilted wheel the other night. Max bet was like $1 per marker and one marker per number, which is kind of a limit for not needing to have special permit to run Casino where I live. This explains the lack of competence and quality. However, the wheel was not a toy, but seemed quite well made. Only incompetently installed. I was quite surprised! I watched about 30-40 spins and every time except 4 times did the ball hit the same diamond! If we call that diamond number 8 (vertical), then three times the ball hit diamond number 7 (horisontal), one time diamond number 6 (vertical) and no time any of the other 5 diamonds.
On such a wheel one only needs to estimate the number of rotations the ball has left and how the wheel will stand at that time. Even without any devices, one should be able to target at leaast the right half of the wheel, which in the long run means that all the money bet are doubled. Of course, no real money can be made with $1 bets which explains how this can be happening.
Noone cares to exploit this tilt $100 or so per evening. However, with a prediction device, one could maybe even earn like 4 times the money (with odds 1/9). Almost only the scatter is the limit for precision. One could make $500 per night on such a low-end roulette place. Definitely worthwhile as a practice ground for developing skills and technologies! Most interesting is however that I never imagined that a tilt could generate such a very regular outcome! Interesting also that they changed croupier and still the same results came up. To me, this confirms that the physics is quite simple after all and that it doesn’'t matter how the ball is thrown by the croupier. This suggests to me that roulette should be quite predictable even with non-tilted wheels.
Max bet was like $1 per marker and one marker per number,
Is that meaning $1 max bet per number?
It is still not bad if you play each spin 18 numbers and if you manage 1:18 it is win $18 per spin.
If you define wrong revolution you will be wrong by t x rotor numbers per sec. Also even if you are in same revolution the ball will not travel with same speed remaining revolutions. Therefore time from prediction until impact with diamond can be different therefore rotor may not be in expected position.
It can be difference close to 11 numbers.
Example
Ball speed 1 sec per rotation at your observation point
Ball travels another 4 rotations until hits diamond.
But and ball of 1.1 sec per rotation will travel same distance
However time of traveling will be longer.
If it is longer by 0.8 sec then rotor will not be in same position as in previous scenario. 0.8 x rotor numbers per sec
It is fact, which none of visual players would accept. So someone is wrong.
That calculation I include in my tilt system and no matter with which speed the ball is clocked I get exact number under the diamond as it was predicted.
I get one pick point they have 2 or 4 on same kind of wheel.
To achieve high hit rate such as 1:6 -1:12 on Huxley wheel is possible only with selected spins, and strong tilt as you described.
If scatter is not too wide you should be ok.
The worst is slightly tilted wheel, not much advantage if we apply prediction for tilted wheel or we need to play only selected spins where we know exactly where the ball will hit diamond or which diamond, if we apply level system on slightly tilted wheel we lose on accuracy because of tilt.
It is fact, which none of visual players would accept. So someone is <<wrong
Oh we accept that of course. But it will show in our tracking. If we frequently have a large difference in the remaining running time, there is usually either 2 peaks located not far from each other or a wider peak covering a 5-9 number arc. Which naturally only makes things easyer.
Or with the sweet speed: No difference at all.
Its very rare to see a very sharp peak located on only 1-3 numbers on todays wheels.
Is that meaning $1 max bet per number?Yes it is! However, one can bet around that number at two, three and four numbers at a time, each including the targeted number. I think it's called to "lay a carpet" of marks on the table.
It is still not bad if you play each spin 18 numbers and if you manage 1:18 it is win $18 per spin.It's a win of $36 every 18 spin if one plays on one number at a time, isn't it? Or 36 won, 18 lost if one bets on 18 numbers at a time. That is, $2 return per $1 bet in the long run.
Also even if you are in same revolution the ball will not travel with same speed remaining revolutions.Yes, that complicates things for intuition. I'm not very skilled in such spatio-temporal real time judgement. I didn't try it even on that very sick wheel, and I don't think that is a good way to go for me. I need technical aid!
To achieve high hit rate such as 1:6 -1:12 on Huxley wheel is possible only with selected spins, and strong tilt as you described.Please, how can I identify the type of wheel?
If scatter is not too wide you should be ok.It really didn't seem to be wide at all. Seemed to land about 1/4 of the wheel forward from the diamond. But I did not keep record of it.
I’m actually a bit chocked about how skewed the game was. Even if the low betting limit gives the gamblers only small edge, it is potentially disastrous for the roulette company which must pay salary and other costs. I suppose that it just shows how very rare it is with gamblers who even try to exploit defect roulette tables (where I live). The profit on all the drunk people and the croupier is mainly (if at all) trained in keeping track of the markers so that people don’t put them in their pockets. Well, we do have a “real” casino in town too. I’ll go there one day soon…
The worst is slightly tilted wheelYes, that's too bad isn't it, because most wheels are slightly tilted, aren't they? Perfection is almost inachievable and big tilts are quickly discovered (where bigger money is involved).
However, with statistics collected on a specific wheel, it should in principle not matter how it is tilted. If it isn’t perfectly random (and even a perfect wheel isn’t perfectly random given that we can estimate the speed of the ball and the wheel at some point in time), then there is a pattern to profit from. The same pattern should repeat itself until the wheel is changed. It does require both dynamic modelling and quite a sample though.
'Good prediction on tilted wheel could partly predict ball scatter as well, sure there may be some unexpected bouncing across the wheel but some can be predicted. By knowing exact ball speed it is possible to know where the ball will hit diamond or will it hit particular one. If ball going towards very top part of diamond, traveling of that ball after diamond is not very predictable. If we play wheel where 70% of ball speeds from fifth revolution before the end is grouped because of tilt we end up with only 50% highly predictable spins. Sorry Rollo I do not know how identify wheels, I do it my way. Same looking wheel may be very different so why to bother about model.
I do care about deceleration that I define on particular wheel / ball combination. I do not play scalloped wheel. Not many players are taking advantage of tilted wheel. Recently I come across one; it was there for about 4 months. I could never see anyone taking advantage of that wheel except myself. On the end casino levelled the wheel and changed NMB to much earlier time. Yes, most of wheels are slightly tilted; few days ago I watched game. Around table there were probably 15 people. On the end there was a huge man of about 140 kg of weight. To reach numbers he would support his weight on dealer side of table. At that moment all table wobbles and it looked to me that wheel moves by 10 mm. I do not know how to get advantage on tilted wheel without timer or computer calculation. Actually I would be more correct if I say that I never applied any of visual techniques on tilted wheel, except chasing dealers signature. Kelly claims to be an expert in the field and that he has experience to notice and act according small difference in wheel speed change of better then 0.1 sec. I hardy can see difference if wheel takes 3 sec or 3.5 sec per revolution.
gkd wrote :
[i]I note your reference to the timing error problem inherent in most systems where data is collected manually.
My backgound is in designing ICs that detect patterns in data in real time. I am planning a project to create a hardware solution to measuiring wheel/ball timings without manual intervention - hence my visit to your site. When I’ve made some progress I’ll post news.[/i]
Along these same lines, has anyone tried producing a prediction solution using neural nets software such as NeuroSolutions by NeuroDimension? It would seem that this kind of software should be able to come up with something useful.
anam
I don`t think i ever claimed to be an expert.
An example from the music world: Wouldnt you expect a skilled drummer to be able keep a specific tact going, like: tap, tap, tap with a certain interval of say 1.0 sec ?
Now if zero passes 12 oclock at 1st tap and then passes 12 oclock at third tap, you have a perfect 3.0 rotor.
If zero has passed with 1 number at the third tap, you have a rotor that is at ~2,95 depending on the acuracy. And if zero is still one number from 12 oclock at the third tap you have a ~3,05.
On slow rotor speeds you can do it using only 1/2 a wheel rotation (another option is using 2 taps) but if you got the time, then use a full rotor rev.
Finally, the vb players is also using a marching pattern after the prediction. Meaning the number must be at a certain spot in the revoloution after the prediction. If it is not, then you either correct the prediction or if the number was too far away from where it should be, discard the spin as a spin you have misjudged.
I do not know how to get advantage on tilted wheel without timer or computer calculation
Actually, if you do a analyse of Scotts training videos where the wheel is level, slightly tilted and really tiltet, (rotor speeds between 2.2 - 2.6 sec) you will at some speeds/and degrees of tilt have 11 numbers arcs with 76% edge and SD between 3.02 - 4.02 using either cross over patterns or acoustic method.[/quote]
I’m beginning to wonder if there is a general solution to roulette prediction, or if we are relegated to specific solutions to particular wheel/ball combinations.
Yet, there may still be a general approach that may merely need to be ‘tweeked’ for a given wheel/ball.
Comments?
anam
I looked for post on GG where Mark was analysing Laurence’s system. It is not there any more or I could not find it. I do have Laurence’s book and I do not find it practical. To define right revolution by using rotor as reference where only 0.1 sec difference can indicate wrong revolution not many people would be able to do. On such high rotor speed 2.5 sec or 2.6 sec per rotation not only that may indicate wrong revolution. rotor position under the diamond when ball arrives would change by 5-6 numbers as well. So if I try to use it for me rotor of 2.5 or 2.7 sec per revolution or anything in between would look the same.
I think if I slow count 21, 22, that I would be more stable then rotor. At least if practising counting and using it in proper way rotor change can be noticed and with adequate multiplying factor can lead to better prediction.
Forester,
You state Leveled wheel.
Same wheel 1. kind of ball, system learns wheel /ball and predicts according to it.
Same wheel 2. kind of ball, system learns wheel /ball and predicts according to it but predicted number does not have to match final number in same distance as with example with ball 1. It can be done but there is no need for it, because it can cause more problems then benefits.
Are you referring to your particular system, or are you saying that a general solution is not possible.
anam[/u]
Also, Forester…
Do you think Neural Nets (Artificial Intelligence) has any usefulness in roulette prediction?
anam
Along these same lines, has anyone tried producing a prediction solution using neural nets software such as NeuroSolutions by NeuroDimension? It would seem that this kind of software should be able to come up with something useful.Neural networks find relations between input and output data very much like non-linear regression does. I think that they are actually mathematically equivalent. However, while statistics sounds boring, neural networks sounds cool so it's a way to trick students into statistics (and investors to put money on such research)...
It has to do with “curve fitting”. Imagine a diagram where you have plotted samples of measaured speed of the ball on the X-axis against time until ball leaves the ball track on the Y-axis. To make predictions, you want to fit a curve to those plots, a curve described by a neat formula which only needs the measured speed as input in order to give the time until the ball leaves the track as output. Here one could use linear regression (forcing a straight line to fit in), but that would probably give a poor fit and big errors, even though on average it would be correct. Or you could fit a bent curve, that is non-linear regression. Using neural networks should give similar result. Physically, I’ve read that the relation between ball speed V and time T to leaving the ball track, should look like this: T = a*ln[1/(V-c)] so fitting a curve like that, choosing values for a, b and c to make the best fit using sample data, should be optimal.
Separately, one should measure the speed (and deceleration A if it is significant) of the wheel as well and using a simple physical relationship like S = VT (or S = AT^2) to find out the distance S which the wheel will move until time T when the ball starts falling down. Then consider where the probability peaks in the scatter distribution are relative to that position.
This is a theoretic approach. Surely, practical shortcuts can be taken. If you want to find out whether the wheel is tilted or not, then just note what diamonds the ball hits and look for statistically significant skewness in that distribution. Scientific pocket calculators often have prepared functions for that, but some rule of thumb and a certain diamond-hit counting system should be enough too. You only need to use your own neurons for that…