Norm Wattenberger responds
I asked Norm Wattenberger about poison indexes and he responded that, as fas as he can recall or understand, this term would apply for an index that its use could actually bring worse results for the player than not using an index at all!
In this sense, I find the term very interesting, if not (medically) accurate.
"Norm Wattenberger sent me several results indicating where there were some slightly poison indexes, indicated in his simulatons, but none was significant to the standard 3 sigma level. His highest significance was one index, I seem to recall 16 v. 7, Uston +/-, that was 'poison' to the 2.7 sigma level, in 10 Billion hands. There were others, with the strangest being some hit/stand indexes that showed such slight minus outcomes, having strange correlations to rules that there was non rational explanation for."
It is difficult to decipher the above with any confidence, but my understanding of poison, in this context, is a play whose index reverses for a single, specific count level up the ladder, at least once. I too have encountered plays with such indexes way back when I started generating my own tables, using Stanford Wong's early commercial version of BJCA.
...An interesting side note, if not the cause for a minor flame war, would be Norm Wattenberger's opinion about the reliability of the available Index Simulators in the market today. But that is up to Wattenberger to post publicly and not me.