Benjamin Portheault links this tweet as an answer to a similar question under his tweet.
If I understand the metric correctly, then each game is evaluated and for the worst and best positions of that game a win probability is computed.
The waste points are then taking the win probabilities of the best evaluations and compare them with the actually achieved points. Similarly, the save points are computed for the win probabilities of the worst positions and compared against the actually achieved points.
The shark points is then the number of points a player could have achieved for never "wasting" together with their actual "saving".
For example, Vincent Keymer appears to have played many games in which the engine was giving him good evaluations at some point in the game.
However, he did not manage to capitalize on that and scored only 3.5 points.
According to the waste measure, he should have scored another 4.51 points to a total of 8.01.
Magnus Carlsen, conversely, was often in positions that were disfavored by the engine.
Yet, he scored a total of 6 points, which is 4.06 points more than predicted according to the save measure.
As the all these measures are computed as the sum over probabilities, they do not correspond to actual game outcomes (as can be seen by the decimal values) and it could be that two players would gain shark points for the same game. For example, if Keymer plays Carlsen and first Keymer reaches a position that is strongly evaluated in his favor, and later Carlsen has the better evaluation, but in the end they draw.
In this case, both players would have "wasted" this game and would have corresponding waste and shark points.
Yet, realistically, only one player could have turned their advantage into an actual score point.