Sign in

Blog · Coverage and territory · Trading

Axe hit rate on ten axes: the whole arithmetic on one page

The complete axe distribution calculation on ten axes sent by one desk, small enough to check by hand: each axe's instrument, side and the clients it was sent to, the blotter's trades in those instruments within the window, the match rule, axes received and traded per client, the axe hit rate, the client on every list who traded none, the client who traded the desk's axed instrument without being sent the axe, the salesperson's list age, and the assertion that matched trades are a subset of the blotter, so a reader can reproduce every figure and then run it on their own axe log.

The short answerTen axes over a week, each sent to a distribution list of clients, each in one instrument on one side. The blotter's trades in those instruments on the opposite side within four hours are matched to the axe by client. Client X received all ten and traded none; client Y received four and traded three; client Z received none and traded two of the axed instruments on the right side in the window. Axe hit rate per client is traded over received; Z's two trades are traded-unsent. Matched trades are flagged in the blotter, never added to it, and the salesperson's list, unchanged in fourteen months, hits 12 percent. Every number can be reproduced by hand.

Axe hit rate is a join between the axe log, the distribution records and the blotter, and on ten axes it can be done by hand. This page works the sends, the matches, the rates per client, traded-unsent, the list age and the assertion. Window: 4 hours. Product: credit.

The axes and the distribution

Axe Instrument Desk side Sent Sent to
1 Bond A Sell Mon 09:00 X, Y, W
2 Bond B Buy Mon 10:30 X, W
3 Bond C Sell Mon 14:00 X, Y
4 Bond A Sell Tue 09:15 X, W
5 Bond D Buy Tue 11:00 X, Y, W
6 Bond E Sell Wed 09:00 X
7 Bond B Buy Wed 13:30 X, Y
8 Bond F Sell Thu 09:00 X, W
9 Bond C Sell Thu 15:00 X, W
10 Bond D Buy Fri 09:30 X, Y, W

The blotter, trades in the axed instruments during the week

Trade Client Instrument Client side Time Matches axe?
T1 Y Bond A Buy Mon 10:10 Axe 1: yes
T2 Z Bond B Sell Mon 11:45 Axe 2 window, right side, not sent: traded-unsent
T3 Y Bond C Buy Mon 16:30 Axe 3: yes
T4 W Bond A Buy Tue 10:00 Axe 4: yes
T5 Y Bond D Sell Tue 13:20 Axe 5: yes
T6 Z Bond D Sell Fri 11:00 Axe 10 window, right side, not sent: traded-unsent
T7 X Bond G Buy Wed 10:00 No axe in Bond G: unrelated
T8 W Bond F Buy Thu 16:00 Axe 8: outside the 4-hour window; no match

The assertion

matched trades ⊆ blotter: T1, T3, T4, T5 are blotter trades, flagged. Nothing added. Holds.

Axes received and traded, per client

Client Axes received Axes traded Hit rate Traded unsent
X 10 0 0% 0
Y 5 (1, 3, 5, 7, 10) 3 (1, 3, 5) 60% 0
W 7 (1, 2, 4, 5, 8, 9, 10) 1 (4) 14% 0
Z 0 0 2 (T2, T6)

The two lists

Receives everything, trades nothing: X. Ten axes, no trades. Remove from the default list or ask why.

Trades the axes, not sent them: Z. Two of the desk's axed instruments, right side, in the window. Add Z.

The salesperson's list

Salesperson List Size Last changed Axes sent through it Traded List hit rate
S-1 Credit default X, Y, W 14 months ago 22 client-sends 4 18%

Fourteen months without a change, and the client who trades the axes is not on it.

Where it goes wrong, even at ten

Axes measured by sends. X received the most; the list looks busy.

Blotter not joined. Nobody knows Y trades three in five.

Traded-unsent unread. Z stays off the list.

T8 counted. Outside the window; a trade five hours later may be another conversation.

From ten to ten thousand

The same join per axe, per client per product, the same two lists, every week. Covirage runs it on the axe log, the distribution records and the blotter. The axe distribution guide covers the measure, and the hit ratio worked example covers the tiers the lists should be built from.

Questions people ask

What is the match rule?

Same client, same instrument or a stated equivalent, opposite side to the desk's axe, within a stated window after the axe was sent, here four hours. A trade that could match two axes goes to the more recent. The rule is stated; the confidence is shown.

Why is Z the strongest signal?

Because Z took the other side of the desk's axed instrument twice, in the window, without being told about the axe. Z would have traded the axe. Z is not on the list, and X, who has never traded one, is on every list.

How is the list age read?

The distribution list's last change date. A list unchanged for fourteen months was built on who was asked then, not who trades now. Its hit rate beside its age is the salesperson's review.