It's pretty incredible how fast you can get challenged as people try to find chinks in the armor of complete second order risk. Within a day of posting our last piece, we were asked to prove the value in a real life extreme day. Luckily, with the geopolitical instability, May 29th of this year proved to be an ample challenge.
And it makes sense. 7y rate had a 5 standard deviation move (Normalized to the 1m7y implied volatility). That's one heck of a challenge and even we were a bit hesitant, because there will be significant third order effects on a day like that. Frankly, no matter how incredible we feel the PLATSON system is, we aren't magicians or fortune tellers. Our methods are rooted in a better mathematical understanding of the structure of derivatives and markets.
Of course, at PLATSON our approach to such challenges is not to just shift+F9 normally. But we decided to follow the 25 year old adage, "update your curve, update your vols, shift F9 you're a trader". It turns out, just doing it and analyzing later is far superior to trying to predict complex behavior. So we input our test portfolio and started on the challenge.
So in turn, update the curve and vol:
And Shift+F9 to get actual delta and vega change, and PnL:
While the curve moves are extreme at the bucket level, the net isn't nearly as bad as I expected. But that Vega change is scary. Let's remember what we were afraid of on May 29th: a crippling anti-EU government gaining power in Italy and bringing down the EU from the inside. Imagine waking up, or even going home that evening, realizing you were 200k shorter vega just as the world was heaving?
In fact even scarier is how different this vega change is from what saw in our previous run. The location of the risk is completely different, with significant changes in the gamma section and 1y-3y expiries. There is very limited long dated vega risk changes. And had you used the insight you gained from our previous scenario to hedge this day, you'd probably end up with a Texas hedge.
So what was the outcome? Delta Errors:
The Risk errors are significant, and of course waiting till the next day, when the move could have reversed, would mean an extraordinary missed opportunity. 640k is on the line.
How did PLATSON perform?
15k of backend risk. This contains the missed opportunity risk from 640k to only 240k. We think this 400k is enough to sit up and take notice.
And how did PLATSON handle the vega changes?
16k overall, with the worst performance in the 3m and 6m sector, worth about 6k of vega across several buckets. PLATSON successfully predicts 92% of a truly extreme Vega change. If a trader caught even just 0.25 bp vega, this would be worth 50k. In reality, the changes offer much more significant trading opportunities in the volatile environment.
PnL performance tracks the above results:
The nonlinear PnL prediction had an error more extreme than the actual PnL for traditional metrics! That's a stunning outcome for the book and indicative of how useless current risk metrics can be when things start to move.
Platson, on the other hand, has significantly smaller errors (~73% reduction). Add the ease of explaining an 80k break on a 16 basis point move day, along with the far superior risk prediction, and experienced traders can see the value of full risk just for peace of mind.
Reach out today and find out how PLATSON can significantly improve your risk management and trading efficiency. Our system is straight forward, provides well understood Taylor Series risk metrics, and can be implemented quickly for market standard models.
Comentários