But I should state how that I was able to discern the rounding error in Norm Wattenberger's post on his sims, that was being used to support the theories oon FA, where I was chased off for daring to challenge them. Try it yourself and you can isolate the bow effect elements from the rouding errors.
The url posted and linked by Norm was qfit.com/fal.jpg Start with the equation I gave that any possible true count has to match: N*52/(# of cards left). Follow the line along the TC=+1 line to where it peaks and vanishes. Every valid report on true counting will have such a peak where at some number of cards left that true count never appears. If the rounding is to the nearest half true count, then TCs of +1 vanish when less than 52/1.5 cards are left. If the rounding for reporting is to the range +1 to +2, then TCs of +1 vanish at the 26 cards left point.
That establishes the rounding used.
Now go over the peak line. Notice that the rotation of the 3D chart obscures the display past this peak line. Pity because then it would have been clear that certain true counts are not always possible, and the slope of the drift within the rounding range would have been clear. Then it would also have been possible to determine if the highest TC displayed, TC=+10, was ever shown to be rolling off as per the bow effect at some point. All that appears is this peak line, which is a little before the exclusion depth, but the view is not the best. The numberic charts are at bit massaged to match how the graph is displayed and don't add much info to these areas.
It is suspicious in a 30 Billion hand simulation that no hands at the high ranges are graphed until the penetration is fairly deep. Comeon there must have been a few rounds?
Norm it is a great chart and great software, but the views and rounding chosen does not really give as much information as is needed to isolate the data that actually matters. What matters is what happens to the edge when the possible TCs drift upward in a range of TCs used for segmenting the data and being sure to compare equivalent true counts. The final range where you can have an actual given TC showup is a good checkpoint.
