So, by now we all know it - there is no 750 GeV resonance in LHC data. But will we ever learn the lesson ?
The facts

Let me start this post by recalling the bare facts, and a quick-and-dirty introduction for anybody who has been on the Moon in the last eight months or so. Last December, the CMS and ATLAS experiments at the CERN LHC collider presented in two back-to-back seminars their first results on data collected at unprecedented proton-proton collision energy of 13 TeV. The 60% higher center-of-mass energy with respect to collisions analized in the previous years left hopes alive for the discovery of some new physics process, which could have been hiding until then thanks to the large required energy to turn on the reactions. 

At the very end of both talks, given by Marumi Kado (ATLAS) and Jim Olsen (CMS) in the CERN auditorium, the slides contained news about some funny and coincident excesses of events containing photon pairs. The mass spectra could be taken as an indication that there existed a particle with a mass of 750 GeV, which decayed to two photons. 

The statistical significance of the effect was in the 3-4 sigma ballpark in each experiment. In other words, a signal at least this striking could only appear in each experiment only once in a few thousand times. But a coincident signal in the two experiments was much more improbable than that, as the two experiments analyzed their data independently.

The announcement generated a wave of interest that materialized in many newspaper articles and in over 500 theoretical papers investigating possible explanations of the observation. That hype has only ended today, as the 4-5 times larger datasets collected in 2016 do not confirm the 2015 find.

My commentary

So, after day 2 of the ICHEP conference, which is taking place in Chicago this year, everybody has gotten the news. There is no new physics at arm's reach, in the guise of a improbable 750 GeV boson decaying in the improbable diphoton final state. 

Yes, the boson was improbable. Don't trust theorists, who always have an explanation for everything in the drawer - they must obviously have very big desks with lots of drawers. Nobody had ever dreamt of betting a dime on that particular way new physics would be discovered, before the craze of last December begun. And yes, its decay was improbable, too - indeed, until last December no LHC jamborees unveiling results of data analysis in new energy ranges had focused on that signature for exotic searches, leaving much more space to dileptons and dijets of high energy, or missing Et plus jets, for instance.

Regardless, you have seen the story as it developed: a lot of media hype, many genuine and interesting results overshadowed, and a rush to publish theory interpretations and get dozens of citations just for having taken a spot around the table. Even the most respectable theorists abandoned all their other occupations and got in the game of instant publishing whatever new physics theory could contemplate the alleged signature. 

Don't get me wrong here: theorists are not to blame too much - that's their job to jump on any hint of new physics, and although there are many things to criticize about 500+ papers appearing out of the blue in six months, discussing over and over again the same concept (you can add a boson to your theory in this or that way, and you can make it show up as a photon pair first, without paying a large price in terms of coherence of the construct), my take today it really that we ought to search for the source of trouble in the experimentalists' field.

First and foremost: one day I used to argue that a 3-sigma signal is nothing, as effects which can occur by chance only once every 1000 times do happen, if you do thousands of searches (and we do). But today, we need to reckon that not even 4 sigma are enough. A four standard deviation effect arises by chance in a single observation less than once in ten thousand times ( to be precise, once in thirty thousand times). But we almost never do single-observation searches: we do searches for composite hypotheses, which entail free parameters that can take on hundreds of values. And we do hundreds of such searches. 

So you do the math, knowing that the probability of independent effects which are individually rare add linearly together...

I can hear an objection here. Many have gotten engrossed by the 750 GeV thing because both ATLAS and CMS had seen the same effect, albeit with different strengths. Now, this is not entirely correct, as the genesis of the coherent find was a bit different; experiments nowadays talk to each other at the higher management levels, when there is something worth mentioning. This in my mind is a bit like doing blind versus non blind analyses. 

But more important is in my opinion the fact that both experiments gave a lot of relevance to the diphoton excesses, putting them at the end of their jamboree talks. And even before the talks, the fact that something funny was going on in diphoton spectra had been propagated through the community even outside CERN. It is this wave of overdue attention that amplified the reaction of the press and of the theorists. Should we, as experimentalists, be more careful in similar circumstances?

I have been arguing for a long time that, as Wilde once claimed, it is important if people talk about you, even if they talk bad. If HEP receives some overhyping it's not damaging, in my opinion. But since my opinion has always been a minority one, I wonder what people think about the 750 GeV thing today. And I wonder if there will be some serious debate on how to handle similar effects in the future.

This is all I have to say about the 750 GeV resonance. But let me add one small bit of information on a different topic at the end of this post now.

Yesterday, also gluino search results have been shown at ICHEP. And the gluino mass, in most SUSY instantiations, cannot be smaller than 1.9 TeV according to experiments. So we are far above the "psychological" threshold of 1.5 TeV that some theorists had set for this particle. If we are to retain a minimum of coherence, we have to say it loud: SUSY is not a preferred new physics theory anymore. What should we replace it with, though, I haven't a clue. For sure not with a diphoton resonance, anyway.

Below is the exclusion mass plane for gluinos, fresh out of the press. On the x axis is the hypothetical gluino mass, on the y axis the neutralino-1 mass. ATLAS search, 2016 data, 13.3 inverse femtobarns.