The compelling case for Liel Abada's Celtic reinstatement

Maintaining objectivity can be difficult, as we all have biases and many prefer to avoid embarrassment or the feeling that goes along with being publicly wrong. However, a robust analytical process can offer solace if accuracy is the ultimate goal, embarrassment and feelings be damned.

With that introduction out of the way, it is time to come clean on the initial analysis I did on Celtic winger Liel Abada, where I believe the weight of the evidence no longer supports the conclusions drawn.

The initial analysis was similar to the benchmarking exercise I have been doing this summer for various recruitment links and signings. The findings on Abada at the time proved to be accurate in a way, but ultimately wrong.

The Israeli has struggled with many technical aspects of playing at Celtic, but the good has outweighed the bad by some distance. 

To place this into some perspective, the first radar compares Abada and Jota’s output in all league minutes since their respective arrivals:

Obviously, Jota is a high benchmark with which to compare but Abada more than held his own. While the sample size is limited for both, the following radar compares the two against the higher-quality competition (excluding Abada’s single game versus FK Jablonec) in last season’s Europa League group games:

Once again, Abada held his own compared to Jota’s output per on-ball value (OBV). The introduction of OBV by StatsBomb occurred after the initial analysis conducted on Abada, and part of building and maintaining a robust analytical process is to improve it when possible.

The model design of OBV is tailor-made for a player like Abada, who has a lot of positive and negative actions as part of his game. 

For example, we can see from the first radar above that Abada’s ratio of turnovers to open-play passes was about 10.5 per cent, which was significantly higher than Jota’s 8.4 per cent.

If it has felt like Abada has turned the ball over at a healthy rate when he touches the ball, in this instance that feeling aligns with his data. What OBV offers is a framework to try to measure whether his attacking output has offset this deficiency. As we can see from his passing OBV percentile, this is evidence it has. 

This more conventional radar shows attacking-related metrics such as expected goals (xG):

Here we see more clearly the components of good and bad from Abada, as his xG was very high, but he was also quite wasteful when dribbling and crossing, which combined with the rate of turnovers per pass, introduces a reasonable question...

How can a winger playing at a club like Celtic have below-average success rates in those three facets of the game and still be a valuable player?

The answer boils down to quality over quantity. When Abada does succeed with the ball, really good things have happened - good enough to overwhelm when they do not.

For example, the average xG assisted per key pass of 0.15 and xG per shot of 0.17 were both significantly higher than the team average xG per shot of 0.12 over the same period. For context, Jota’s averages have been 0.17 and 0.10 in those metrics. 

This is an example of how I believe the addition of the OBV model has made the analytical process more robust.

It also introduces an interesting question as it relates to player selection relative to Celtic’s front three: does Abada offer enough performance value to make him a reasonable alternative to supplant Daizen Maeda? 

With the caveat that Maeda’s sample remains modest and the extenuating circumstances surrounding his transfer and transition to Scotland last season are important context, his performance output to date has been dwarfed by Abada. 

What has stood out for Maeda has been his relatively low volume of involvement, though him playing about 38 per cent of his league minutes to date at striker should be considered.

Similar to Abada, he has struggled to beat opponents via the dribble and his crossing accuracy has been woeful. However, his ball progression has been extremely weak and his overall OBV has been quite low.

Maeda’s calling card and reputation from his time in Japan was his defensive work rate and pressing and so far his activity has been very high. An issue for Maeda may be that Celtic simply does not have to do that much defending domestically.

That has been reflected by his defensive action OBV having been just 0.02 to date. In addition, his 'Helter Skelter' pressing style has to date made him relatively easy to bypass via the dribble, with him 34th percentile in that metric.

The first two league games this season may be a precursor for what is to come, with Celtic largely in control, dominating possession and creating far more chances than the opposition.

The ability to create and finish quality chances against bunkered and overwhelmed opponents may have a direct impact on the league race.

At least in league games versus 10 teams, Abada may offer a more compelling risk and reward profile than Maeda in this regard. It will be interesting to see if Ange Postecoglou eventually comes to a similar conclusion.