Surprise (Not): Madoff’s Results Were Generated Via a Randomizer

I’m so clever. Back in December I wrote here that Bernie Madoff’s results, far from being pulled from his hat, showed every sign of having been generated by a randomizing algorithm. My analysis suggested that – based on Benford’s Law fit – his results would have passed muster in terms of looking sufficiently random to be real.

Here is the original chart again:

Now, let’s flip forward to today. We see the following in a new SEC complaint against two of Bernie’s programmers:

Defendants knew that all of this data was, for all intents and purposes,
randomly generated and assigned to purported trades by programs they created.
Defendants knew that this was unlike the [internal] House 5 system, on which trade blotter
came from actual executed trades and was verified through data received from
counterparties and clearing agencies. Some of the House 17 programs even included
randomness checks, i.e., code that analyzed program results to ensure they were
sufficiently random

Defendants’ extensive use of random selection arose from two related
facts. First, there were no real trades from which to draw actual data. Second, Madoff
and DiPascali were concerned that investor representatives and/or regulators would
closely examine the data and notice implausible correlations ~, all fake trades with a
certain counterparty were for a certain share volume, or were executed at uniform time
intervals), which could lead to greater scrutiny. In fact, when Madoff, DiPascali, or
Defendants reviewed the programs’ results and found suspicious correlations and
patterns, Defendants would have to make further revisions to the programs.

[Emphasis mine]

Surprise, surprise. Or not.