"Your defensive metrics this month are the best of your career," Mars says, measuring protein powder with the precision of a pharmacist.
"Thank you."
"The improvement correlates with a change in your off-ice behavior that began approximately five weeks ago."
I add ice to my blender. The ice cubes are loud against the steel.
"I track correlations," Mars says. "It is compulsive. It is not something I can turn off. Your defensive positioning has improved by six percent. Your reaction time on gap control has decreased by a tenth of a second. Your communication with your D-partner has increased in both frequency and accuracy. These are measurable improvements. The cause of the improvement is also measurable, though less conventionally."
"Mars."
"Mercer."
The name sits between us on the smoothie bar counter. Mars does not look at me when he says it. Mars looks at his shake. The not-looking is the courtesy. Mars is offering the name without the eye contact, the data without the accusation, the seeing without the demand to be seen back.
"I am not asking you to confirm," he says. "The data is sufficient. I am telling you what the data shows because the data shows something else, and the something else is what I came to say."
He blends his shake. The blender runs for eight seconds. He pours. He drinks. The sequence is executed with the mechanical efficiency of a man who has optimized every physical process in his life, including the consumption of nutrients.
"The improvement is real," Mars says. "The improvement is also fragile. The fragility is not in the hockey. The fragility is in you. You are playing better because something in your life has changed, and the change is making you lighter, and the lightness is making you faster, and the faster is making you better. But the lightness depends on a condition that you are currently managing rather than trusting."
The sentence is Mars's version of Mik's sentence, translated from Mik's language (emotional, experiential, built from elevenyears of hiding) into Mars's language (analytical, predictive, built from a career of reading patterns). Two men. Same observation. Different dialects.
"You are managing Mercer the way you manage a defensive assignment," Mars says. "Proximity control. Gap assessment. Threat evaluation. The managing is producing results because the managing gives you a framework and frameworks are where you feel safe. But Mercer is not a defensive assignment. Mercer is a person. And the framework you are applying to a person is the framework that failed you with the previous person, not because the framework was wrong but because the previous person was wrong."
I set my shake down. The setting-down is deliberate. My hand wants to grip the cup and the wanting-to-grip is the tell and Mars will see the tell and Mars sees everything.
"The previous person," I say.
"I do not know the details. I know the shape. The shape is visible in your behavior: the management, the distance calibration, the way you control proximity as if proximity is a threat variable. This is the behavior of a man who was hurt by someone who was close, and whose response to the hurt was to reclassify closeness as danger."
Mars drinks his shake. The drinking is calm. Mars is calm the way deep water is calm: the surface is still and the depth is enormous.
"The reclassification was logical," he says. "At the time. With the data you had. But the data set has changed. The current person is not the previous person. Applying the old model to new data produces inaccurate predictions. You are predicting a threat that does not exist in the current data."
"How do you know the threat does not exist?"
"Because I watch. I have watched Mercer for five weeks. His behavioral pattern around you is consistent, escalatingin investment, and shows zero indicators of the withdrawal or manipulation patterns that characterize the threat you are predicting. He is not managing you. He is choosing you. The choosing is visible in his body language, his performance metrics, and the fact that he went to the video room before calling his mother on cut day. That is not a data point consistent with threat."
The smoothie bar is quiet. Two men standing with protein shakes, discussing behavioral analysis and defensive frameworks, and the discussion is about love, and neither man has used the word, and the not-using is the Mars way, because Mars does not deal in words that cannot be measured. Mars deals in data. And the data says: Eli is safe.
"The last time I stopped managing, I got destroyed," I say.
Mars nods. The nod is not sympathetic. The nod is the acknowledgment of a data point.
"The last time, the person was not worth the risk. That is a data point about him. Not about risk."
The sentence hits the center of my chest. The hit is clean. The hit is a Mars Santos glove save: precise, minimal, devastating.
A data point about him. Not about risk.
The destruction was not evidence that closeness is dangerous. The destruction was evidence that Alexei was dangerous. The two conclusions look identical from inside the damage. From outside the damage, from Mars's vantage point (the crease, the place where you can see the whole shape of the play), the conclusions are different. One says: never again. The other says: not him again.
Not him. Not never.
Mars finishes his shake. He rinses the cup. He places it in the drying rack with the positional precision of a man who considers dish placement a matter of spatial optimization.
"I do not give advice," he says. "I present data. The data suggests that your current defensive model is optimized for a threat that is not present in the current environment. Continuing to run the model will produce a false positive that results in the loss of something the data indicates is valuable."
